# Bibliography of computer-aided Drug Design

Updated on 7/18/2014. Currently 2130 references

## Docking / Scoring functions

2014 / 2013 / 2012 / 2011 / 2010 / 2009 / 2008 / 2007 / 2006 / 2005 / 2004 / 2003 / 2002 / 2001 / 2000 / 1999 / 1998 / 1997 / 1996 / 1995 / 1994 / 1992 /

## 2014

• Exhaustive docking and solvated interaction energy scoring: lessons learned from the SAMPL4 challenge.
Hogues, Hervé and Sulea, Traian and Purisima, Enrico O
Journal of computer-aided molecular design, 2014
PMID: 24474162     doi: 10.1007/s10822-014-9715-5

We continued prospective assessments of the Wilma-solvated interaction energy (SIE) platform for pose prediction, binding affinity prediction, and virtual screening on the challenging SAMPL4 data sets including the HIV-integrase inhibitor and two host-guest systems. New features of the docking algorithm and scoring function are tested here prospectively for the first time. Wilma-SIE provides good correlations with actual binding affinities over a wide range of binding affinities that includes strong binders as in the case of SAMPL4 host-guest systems. Absolute binding affinities are also reproduced with appropriate training of the scoring function on available data sets or from comparative estimation of the change in target's vibrational entropy. Even when binding modes are known, SIE predictions lack correlation with experimental affinities within dynamic ranges below 2 kcal/mol as in the case of HIV-integrase ligands, but they correctly signaled the narrowness of the dynamic range. Using a common protein structure for all ligands can reduce the noise, while incorporating a more sophisticated solvation treatment improves absolute predictions. The HIV-integrase virtual screening data set consists of promiscuous weak binders with relatively high flexibility and thus it falls outside of the applicability domain of the Wilma-SIE docking platform. Despite these difficulties, unbiased docking around three known binding sites of the enzyme resulted in over a third of ligands being docked within 2\AA} from their actual poses and over half of the ligands docked in the correct site, leading to better-than-random virtual screening results.

• Importance of ligand conformational energies in carbohydrate docking: Sorting the wheat from the chaff.
Nivedha, Anita K and Makeneni, Spandana and Foley, Bethany Lachele and Tessier, Matthew B and Woods, Robert J
Journal of computational chemistry, 2014, 35(7), 526-539
PMID: 24375430     doi: 10.1002/jcc.23517

Docking algorithms that aim to be applicable to a broad range of ligands suffer reduced accuracy because they are unable to incorporate ligand-specific conformational energies. Here, we develop a set of Carbohydrate Intrinsic (CHI) energy functions that quantify the conformational properties of oligosaccharides, based on the values of their glycosidic torsion angles. The relative energies predicted by the CHI energy functions mirror the conformational distributions of glycosidic linkages determined from a survey of oligosaccharide-protein complexes in the protein data bank. Addition of CHI energies to the standard docking scores in Autodock 3, 4.2, and Vina consistently improves pose ranking of oligosaccharides docked to a set of anticarbohydrate antibodies. The CHI energy functions are also independent of docking algorithm, and with minor modifications, may be incorporated into both theoretical modeling methods, and experimental NMR or X-ray structure refinement programs.

## 2013

• Water PMF for predicting the properties of water molecules in protein binding site
Zheng, Mingyue and Li, Yanlian and Xiong, Bing and Jiang, Hualiang and Shen, Jingkang
Journal of computational chemistry, 2013, 34(7), 583-592
PMID: 23114863     doi: 10.1002/jcc.23170

Water is an important component in living systems and deserves better understanding in chemistry and biology. However, due to the difficulty of investigating the water functions in protein structures, it is usually ignored in computational modeling, especially in the field of computer-aided drug design. Here, using the potential of mean forces (PMFs) approach, we constructed a water PMF (wPMF) based on 3946 non-redundant high resolution crystal structures. The extracted wPMF potential was first used to investigate the structure pattern of water and analyze the residue hydrophilicity. Then, the relationship between wPMF score and the B factor value of crystal waters was studied. It was found that wPMF agrees well with some previously reported experimental observations. In addition, the wPMF score was also tested in parallel with 3D-RISM to measure the ability of retrieving experimentally observed waters, and showed comparable performance but with much less computational cost. In the end, we proposed a grid-based clustering scheme together with a distance weighted wPMF score to further extend wPMF to predict the potential hydration sites of protein structure. From the test, this approach can predict the hydration site at the accuracy about 80% when the calculated score lower than -4.0. It also allows the assessment of whether or not a given water molecule should be targeted for displacement in ligand design. Overall, the wPMF presented here provides an optional solution to many water related computational modeling problems, some of which can be highly valuable as part of a rational drug design strategy.

• Evaluation and Optimization of Virtual Screening Workflows with DEKOIS 2.0 - A Public Library of Challenging Docking Benchmark Sets.
Bauer, Matthias R and Ibrahim, Tamer M and Vogel, Simon M and Boeckler, Frank M
Journal of chemical information and modeling, 2013, 53(6), 1447-1462
PMID: 23705874     doi: 10.1021/ci400115b

The application of molecular benchmarking sets helps to assess the actual performance of virtual screening (VS) workflows. To improve the efficiency of structure-based VS approaches, the selection and optimization of various parameters can be guided by benchmarking. With the DEKOIS 2.0 library, we aim to further extend and complement the collection of publicly available decoy sets. Based on BindingDB bioactivity data, we provide 81 new and structurally diverse benchmark sets for a wide variety of different target classes. To ensure a meaningful selection of ligands, we address several issues that can be found in bioactivity data. We have improved our previously introduced DEKOIS methodology with enhanced physicochemical matching, now including the consideration of molecular charges, as well as a more sophisticated elimination of latent actives in the decoy set (LADS). We evaluate the docking performance of Glide, GOLD, and AutoDock Vina with our data sets and highlight existing challenges for VS tools. All DEKOIS 2.0 benchmark sets will be made accessible at http://www.dekois.com .

• Grid-based molecular footprint comparison method for docking and de novo design: Application to HIVgp41
Balius, Trent E and Allen, William J and Mukherjee, Sudipto and Rizzo, Robert C
Journal of computational chemistry, 2013, 34(14), 1226-1240
PMID: 23436713     doi: 10.1002/jcc.23245

Scoring functions are a critically important component of computer-aided screening methods for the identification of lead compounds during early stages of drug discovery. Here, we present a new multigrid implementation of the footprint similarity (FPS) scoring function that was recently developed in our laboratory which has proven useful for identification of compounds which bind to a protein on a per-residue basis in a way that resembles a known reference. The grid-based FPS method is much faster than its Cartesian-space counterpart, which makes it computationally tractable for on-the-fly docking, virtual screening, or de novo design. In this work, we establish that: (i) relatively few grids can be used to accurately approximate Cartesian space footprint similarity, (ii) the method yields improved success over the standard DOCK energy function for pose identification across a large test set of experimental co-crystal structures, for crossdocking, and for database enrichment, and (iii) grid-based FPS scoring can be used to tailor construction of new molecules to have specific properties, as demonstrated in a series of test cases targeting the viral protein HIVgp41. The method is available in the program DOCK6.

• Automated Large-Scale File Preparation, Docking, and Scoring: Evaluation of ITScore and STScore Using the 2012 Community Structure-Activity Resource Benchmark
Grinter, Sam Z and Yan, Chengfei and Huang, Sheng-You and Jiang, Lin and Zou, Xiaoqin
Journal of chemical information and modeling, 2013, 53(8), 1905-1914
PMID: 23656179     doi: 10.1021/ci400045v

In this study, we use the recently released 2012 Community Structure-Activity Resource (CSAR) data set to evaluate two knowledge-based scoring functions, ITScore and STScore, and a simple force-field-based potential (VDWScore). The CSAR data set contains 757 compounds, most with known affinities, and 57 crystal structures. With the help of the script files for docking preparation, we use the full CSAR data set to evaluate the performances of the scoring functions on binding affinity prediction and active/inactive compound discrimination. The CSAR subset that includes crystal structures is used as well, to evaluate the performances of the scoring functions on binding mode and affinity predictions. Within this structure subset, we investigate the importance of accurate ligand and protein conformational sampling and find that the binding affinity predictions are less sensitive to non-native ligand and protein conformations than the binding mode predictions. We also find the full CSAR data set to be more challenging in making binding mode predictions than the subset with structures. The script files used for preparing the CSAR data set for docking, including scripts for canonicalization of the ligand atoms, are offered freely to the academic community.

• Characterizing Binding of Small Molecules. II. Evaluating the Potency of Small Molecules to Combat Resistance Based on Docking Structures
Ding, Bo and Li, Nan and Wang, Wei
Journal of chemical information and modeling, 2013
PMID: 23570305

Drug resistance severely erodes the efficacy of therapeutic treatments for many diseases. Assessing the potency of a drug lead to combat resistance is no doubt critical for designing new drugs or new therapeutic combinations. Virtual screening is often the first step in drug discovery and a challenging problem is to accurately predict the resistant profile of an inhibitor based on the docking structures. Using a well studied system HIV-1 protease, we have illustrated the success of a computational method called MIEC-SVM on tackling this problem. We computed molecular interaction energy components (MIECs) between the ligand and the protease residues to characterize the docking poses, which were input to support vector machine (SVM) to distinguish resistant from nonresistant mutants. More importantly, the method is able to predict resistant profiles for new drugs based on the docking structures as indicated by its satisfactory performance in leave-one-drug-out and leave-drug/mutants-out tests. Therefore, the MIEC-SVM method can also facilitate designing effective therapeutic combinations by combining drugs with complementary resistant profiles.

• Accounting for Conformational Variability in Protein-Ligand Docking with NMR-Guided Rescoring.
Skj{\ae}rven, Lars and Codutti, Luca and Angelini, Andrea and Grimaldi, Manuela and Latek, Dorota and Monecke, Peter and Dreyer, Matthias K and Carlomagno, Teresa
Journal of the American Chemical Society, 2013, 135(15), 5819-5827
PMID: 23565800     doi: 10.1021/ja4007468

A key component to success in structure-based drug design is reliable information on protein-ligand interactions. Recent development in NMR techniques has accelerated this process by overcoming some of the limitations of X-ray crystallography and computational protein-ligand docking. In this work we present a new scoring protocol based on NMR-derived interligand INPHARMA NOEs to guide the selection of computationally generated docking modes. We demonstrate the performance in a range of scenarios, encompassing traditionally difficult cases such as docking to homology models and ligand dependent domain rearrangements. Ambiguities associated with sparse experimental information are lifted by searching a consensus solution based on simultaneously fitting multiple ligand pairs. This study provides a previously unexplored integration between molecular modeling and experimental data, in which interligand NOEs represent the key element in the rescoring algorithm. The presented protocol should be widely applicable for protein-ligand docking also in a different context from drug design and highlights the important role of NMR-based approaches to describe intermolecular ligand-receptor interactions.

• Optimization of molecular docking scores with support vector rank regression
Wang, Wei and He, Wanlin and Zhou, Xi and Chen, Xin
Proteins, 2013, n/a-n/a
PMID: 23504920     doi: 10.1002/prot.24282

This work introduces the support vector rank regression (SVRR) algorithm for the optimization of molecular docking scores. Seven original docking scores reported by two docking software were integrated by the SVRR algorithm. The resulting SVRR scores showed an average of 12.1% improvement (59.5% to 66.7%) in binding conformation prediction tests to rank the correctly computed conformation in the first place, along with 16.7% RMSD improvement (2.5414\AA} vs. 2.1162\AA}) for the top ranked conformations. In compound library screening tests, an average of 46.3% improvement (18.2% to 26.6%) was also observed to rank the correct ligand in the first place. Furthermore, it was shown that SVRR scores trained with different example datasets, using different training strategies, all exhibited exceedingly consistent accuracies, suggesting that the SVRR algorithm is highly robust and generalizable. In contrast, using the same training datasets, traditional support vector classification and regression algorithms failed to comparably improve the accuracy of library screening and conformation prediction. These results suggested that, with additional features to indicate the comparative fitness between computed binding conformations, the SVRR algorithm holds the potential to create a new category of more accurate integrative docking scores. Proteins 2013.

## 2012

• A Force Field with Discrete Displaceable Waters and Desolvation Entropy for Hydrated Ligand Docking.
Forli, Stefano and Olson, Arthur J
Journal of medicinal chemistry, 2012, 55(2), 623-638
PMID: 22148468     doi: 10.1021/jm2005145

In modeling ligand-protein interactions, the representation and role of water are of great importance. We introduce a force field and hydration docking method that enables the automated prediction of waters mediating the binding of ligands with target proteins. The method presumes no prior knowledge of the apo or holo protein hydration state and is potentially useful in the process of structure-based drug discovery. The hydration force field accounts for the entropic and enthalpic contributions of discrete waters to ligand binding, improving energy estimation accuracy and docking performance. The force field has been calibrated and validated on a total of 417 complexes (197 training set; 220 test set), then tested in cross-docking experiments, for a total of 1649 ligand-protein complexes evaluated. The method is computationally efficient and was used to model up to 35 waters during docking. The method was implemented and tested using unaltered AutoDock4 with new force field tables.

• Automatic design of decision-tree induction algorithms tailored to flexible-receptor docking data.
Barros, Rodrigo C and Winck, Ana T and Machado, Karina S and Basgalupp, Márcio P and de Carvalho, André Cplf and Ruiz, Duncan D and Norberto de Souza, Osmar
Bmc Bioinformatics, 2012, 13(1), 310
PMID: 23171000     doi: 10.1186/1471-2105-13-310

ABSTRACT: Background This paper addresses the prediction of the free energy of binding of a drug candidate with enzyme InhA associated with Mycobacterium tuberculosis. This problem is found within rational drug design, where interactions between drug candidates and target proteins are verified through molecular docking simulations. In this application, it is important not only to correctly predict the free energy of binding, but also to provide a comprehensible model that could be validated by a domain specialist. Decision-tree induction algorithms have been successfully used in drug-design related applications, specially considering that decision trees are simple to understand, interpret, and validate. There are several decision-tree induction algorithms available for general-use, but each one has a bias that makes it more suitable for a particular data distribution. In this article, we propose and investigate the automatic design of decision-tree induction algorithms tailored to particular drug-enzyme binding data sets. We investigate the performance of our new method for evaluating binding conformations of different drug candidates to InhA, and we analyze our findings with respect to decision tree accuracy, comprehensibility, and biological relevance. Results The empirical analysis indicates that our method is capable of automatically generating decision-tree induction algorithms that significantly outperform the traditional C4.5 algorithm with respect to both accuracy and comprehensibility. In addition, we provide the biological interpretation of the rules generated by our approach, reinforcing the importance of comprehensible predictive models in this particular bioinformatics application. Conclusions We conclude that automatically designing a decision-tree algorithm tailored to molecular docking data is a promising alternative for the prediction of the free energy from the binding of a drug candidate with a flexible-receptor.

• Can the Energy Gap in the Protein-Ligand Binding Energy Landscape Be Used as a Descriptor in Virtual Ligand Screening?
Grigoryan, Arsen V and Wang, Hong and Cardozo, Timothy J
PloS one, 2012, 7(10), e46532
doi: 10.1371/journal.pone.0046532

The ranking of scores of individual chemicals within a large screening library is a crucial step in virtual screening (VS) for drug discovery. Previous studies showed that the quality of protein-ligand recognition can be improved using spectrum properties and the shape of ...

• Can the Energy Gap in the Protein-Ligand Binding Energy Landscape Be Used as a Descriptor in Virtual Ligand Screening?
Grigoryan, Arsen V and Wang, Hong and Cardozo, Timothy J
PloS one, 2012, 7(10), e46532
doi: 10.1371/journal.pone.0046532

The ranking of scores of individual chemicals within a large screening library is a crucial step in virtual screening (VS) for drug discovery. Previous studies showed that the quality of protein-ligand recognition can be improved using spectrum properties and the shape of ...

• Docking and scoring with ICM: the benchmarking results and strategies for improvement.
Neves, Marco A C and Totrov, Maxim and Abagyan, Ruben
Journal of computer-aided molecular design, 2012, 26(6), 675-686
PMID: 22569591     doi: 10.1007/s10822-012-9547-0

Flexible docking and scoring using the internal coordinate mechanics software (ICM) was benchmarked for ligand binding mode prediction against the 85 co-crystal structures in the modified Astex data set. The ICM virtual ligand screening was tested against the 40 DUD target benchmarks and 11-target WOMBAT sets. The self-docking accuracy was evaluated for the top 1 and top 3 scoring poses at each ligand binding site with near native conformations below 2\AA} RMSD found in 91 and 95% of the predictions, respectively. The virtual ligand screening using single rigid pocket conformations provided the median area under the ROC curves equal to 69.4 with 22.0% true positives recovered at 2% false positive rate. Significant improvements up to ROC AUC

• Docking performance of the glide program as evaluated on the Astex and DUD datasets: a complete set of glide SP results and selected results for a new scoring function integrating WaterMap and glide.
Repasky, Matthew P and Murphy, Robert B and Banks, Jay L and Greenwood, Jeremy R and Tubert-Brohman, Ivan and Bhat, Sathesh and Friesner, Richard A
Journal of computer-aided molecular design, 2012, 26(6), 787-799
PMID: 22576241     doi: 10.1007/s10822-012-9575-9

Glide SP mode enrichment results for two preparations of the DUD dataset and native ligand docking RMSDs for two preparations of the Astex dataset are presented. Following a best-practices preparation scheme, an average RMSD of 1.140\AA} for native ligand docking with Glide SP is computed. Following the same best-practices preparation scheme for the DUD dataset an average area under the ROC curve (AUC) of 0.80 and average early enrichment via the ROC (0.1 %) metric of 0.12 were observed. 74 and 56 % of the 39 best-practices prepared targets showed AUC over 0.7 and 0.8, respectively. Average AUC was greater than 0.7 for all best-practices protein families demonstrating consistent enrichment performance across a broad range of proteins and ligand chemotypes. In both Astex and DUD datasets, docking performance is significantly improved employing a best-practices preparation scheme over using minimally-prepared structures from the PDB. Enrichment results for WScore, a new scoring function and sampling methodology integrating WaterMap and Glide, are presented for four DUD targets, hivrt, hsp90, cdk2, and fxa. WScore performance in early enrichment is consistently strong and all systems examined show AUC > 0.9 and superior early enrichment to DUD best-practices Glide SP results.

• PRL-dock: Protein-ligand docking based on hydrogen bond matching and probabilistic relaxation labeling.
Wu, Meng-Yun and Dai, Dao-Qing and Yan, Hong
Proteins, 2012, 80(9), 2137-2153
PMID: 22544808     doi: 10.1002/prot.24104

Protein-ligand docking is widely applied to structure-based virtual screening for drug discovery. This paper presents a novel docking technique, PRL-Dock, based on hydrogen bond matching and probabilistic relaxation labeling. It deals with multiple hydrogen bonds and can match many acceptors and donors simultaneously. In the matching process, the initial probability of matching an acceptor with a donor is estimated by an efficient scoring function and the compatibility coefficients are assigned according to the coexisting condition of two hydrogen bonds. After hydrogen bond matching, the geometric complementarity of the interacting donor and acceptor sites is taken into account for displacement of the ligand. It is reduced to an optimization problem to calculate the optimal translation and rotation matrixes that minimize the root mean square deviation between two sets of points, which can be solved using the Kabsch algorithm. In addition to the van der Waals interaction, the contribution of intermolecular hydrogen bonds in a complex is included in the scoring function to evaluate the docking quality. A modified Lennard-Jones 12-6 dispersion-repulsion term is used to estimate the van der Waals interaction to make the scoring function fairly 'soft' so that ligands are not heavily penalized for small errors in the binding geometry. The calculation of this scoring function is very convenient. The evaluation is carried out on rigid complexes and 93 flexible ones where there is at least one intermolecular hydrogen bond. The experiment results of docking accuracy and prediction of binding affinity demonstrate that the proposed method is highly effective. Proteins 2012.

• 3D-RISM-Dock: a new fragment-based drug design protocol
Nikolić, Dragan and Blinov, Nikolay and Wishart, David S. and Kovalenko, Andriy
Journal of chemical theory and computation, 2012, 8(9), 3356-3372
doi: 10.1021/ct300257v

We explore a new approach in the rational design of specificity in molecular recognition of small molecules based on statistical-mechanical integral equation theory of molecular liquids in the form of the three-dimensional reference interaction site model with the Kovalenko-Hirata closure (3D-RISM-KH). The numerically stable iterative solution of conventional 3D-RISM equations includes the fragmental decomposition of flexible ligands, which are treated as distinct species in solvent mixtures of arbitrary complexity. The computed density functions for solution (including ligand) molecules are obtained as a set of discrete spatial grids that uniquely describe the continuous solvent-site distribution around the protein solute. Potentials of mean force derived from these distributions define the scoring function interfaced with the AutoDock program for an automated ranking of docked conformations. As a case study in terms of solvent composition, we analyze cooperative interactions encountered in the binding o...

• Molecular Docking Using the Molecular Lipophilicity Potential as Hydrophobic Descriptor: Impact on GOLD Docking Performance.
Nurisso, Alessandra and Bravo, Juan and Carrupt, Pierre-Alain and Daina, Antoine
Journal of chemical information and modeling, 2012, 52(5), 319-1327
PMID: 22462609     doi: 10.1021/ci200515g

GOLD is a molecular docking software widely used in drug design. In the initial steps of docking, it creates a list of hydrophobic fitting points inside protein cavities that steer the positioning of ligand hydrophobic moieties. These points are generated based on the Lennard-Jones potential between a carbon probe and each atom of the residues delimitating the binding site. To thoroughly describe hydrophobic regions in protein pockets and properly guide ligand hydrophobic moieties toward favorable areas, an in-house tool, the MLP filter, was developed and herein applied. This strategy only retains GOLD hydrophobic fitting points that match the rigorous definition of hydrophobicity given by the molecular lipophilicity potential (MLP), a molecular interaction field that relies on an atomic fragmental system based on 1-octanol/water experimental partition coefficients (log P(oct)). MLP computations in the binding sites of crystallographic protein structures revealed that a significant number of points considered hydrophobic by GOLD were actually polar according to the MLP definition of hydrophobicity. To examine the impact of this new tool, ligand-protein complexes from the Astex Diverse Set and the PDB bind core database were redocked with and without the use of the MLP filter. Reliable docking results were obtained by using the MLP filter that increased the quality of docking in nonpolar cavities and outperformed the standard GOLD docking approach.

• Protein-Ligand Binding Free Energies from Exhaustive Docking.
Purisima, Enrico O and Hogues, Hervé
The journal of physical chemistry. B, 2012, 116(23), 6872-6879
PMID: 22432509     doi: 10.1021/jp212646s

We explore the use of exhaustive docking as an alternative to Monte Carlo and molecular dynamics sampling for the direct integration of the partition function for protein-ligand binding. We enumerate feasible poses for the ligand and calculate the Boltzmann factor contribution of each pose to the partition function. From the partition function, the free energy, enthalpy, and entropy can be derived. All our calculations are done with a continuum solvation model that includes solving the Poisson equation. In contrast to Monte Carlo and molecular dynamics simulations, exhaustive docking avoids (within the limitations of a discrete sampling) the question of "Have we run long enough?" due to its deterministic complete enumeration of states. We tested the method on the T4 lysozyme L99A mutant, which has a nonpolar cavity that can accommodate a number of small molecules. We tested two electrostatic models. Model 1 used a solute dielectric of 2.25 for the complex apoprotein and free ligand and 78.5 for the solvent. Model 2 used a solute dielectric of 2.25 for the complex and apoprotein but 1.0 for the free ligand. For our test set of eight molecules, we obtain a reasonable correlation with a Pearson r(2)

• Pose prediction and virtual screening performance of GOLD scoring functions in a standardized test.
Liebeschuetz, John W and Cole, Jason C and Korb, Oliver
Journal of computer-aided molecular design, 2012, 26(6), 737-748
PMID: 22371207     doi: 10.1007/s10822-012-9551-4

The performance of all four GOLD scoring functions has been evaluated for pose prediction and virtual screening under the standardized conditions of the comparative docking and scoring experiment reported in this Edition. Excellent pose prediction and good virtual screening performance was demonstrated using unmodified protein models and default parameter settings. The best performing scoring function for both pose prediction and virtual screening was demonstrated to be the recently introduced scoring function ChemPLP. We conclude that existing docking programs already perform close to optimally in the cognate pose prediction experiments currently carried out and that more stringent pose prediction tests should be used in the future. These should employ cross-docking sets. Evaluation of virtual screening performance remains problematic and much remains to be done to improve the usefulness of publically available active and decoy sets for virtual screening. Finally we suggest that, for certain target/scoring function combinations, good enrichment may sometimes be a consequence of 2D property recognition rather than a modelling of the correct 3D interactions.

• Rapid and Accurate Prediction and Scoring of Water Molecules in Protein Binding Sites
Ross, Gregory A and Morris, Garrett M and Biggin, Philip C
PloS one, 2012, 7(3), e32036
doi: 10.1371/journal.pone.0032036.t006

Water plays a critical role in ligand-protein interactions. However, it is still challenging to predict accurately not only where water molecules prefer to bind, but also which of those water molecules might be displaceable. The latter is often seen as a route to optimizing affinity of potential drug candidates. Using a protocol we call WaterDock, we show that the freely available AutoDock Vina tool can be used to predict accurately the binding sites of water molecules. WaterDock was validated using data from X-ray crystallography, neutron diffraction and molecular dynamics simulations and correctly predicted 97% of the water molecules in the test set. In addition, we combined data-mining, heuristic and machine learning techniques to develop probabilistic water molecule classifiers. When applied to WaterDock predictions in the Astex Diverse Set of protein ligand complexes, we could identify whether a water molecule was conserved or displaced to an accuracy of 75%. A second model predicted whether water molecules were displaced by polar groups or by non-polar groups to an accuracy of 80%. These results should prove useful for anyone wishing to undertake rational design of new compounds where the displacement of water molecules is being considered as a route to improved affinity.

• Virtual fragment screening: exploration of MM-PBSA re-scoring.
Kawatkar, Sameer and Moustakas, Demetri and Miller, Matthew and Joseph-McCarthy, Diane
Journal of computer-aided molecular design, 2012, 26(8), 921-934
PMID: 22869295     doi: 10.1007/s10822-012-9590-x

An NMR fragment screening dataset with known binders and decoys was used to evaluate the ability of docking and re-scoring methods to identify fragment binders. Re-scoring docked poses using the Molecular Mechanics Poisson-Boltzmann Surface Area (MM-PBSA) implicit solvent model identifies additional active fragments relative to either docking or random fragment screening alone. Early enrichment, which is clearly most important in practice for selecting relatively small sets of compounds for experimental testing, is improved by MM-PBSA re-scoring. In addition, the value in MM-PBSA re-scoring of docked poses for virtual screening may be in lessening the effect of the variation in the protein complex structure used.

• Flexibility and binding affinity in protein-ligand, protein-protein and multi-component protein interactions: limitations of current computational approaches.
Tuffery, Pierre and Derreumaux, Philippe
Journal of the Royal Society, Interface / the Royal Society, 2012, 9(66), 20-33
PMID: 21993006     doi: 10.1098/rsif.2011.0584

The recognition process between a protein and a partner represents a significant theoretical challenge. In silico structure-based drug design carried out with nothing more than the three-dimensional structure of the protein has led to the introduction of many compounds into clinical trials and numerous drug approvals. Central to guiding the discovery process is to recognize active among non-active compounds. While large-scale computer simulations of compounds taken from a library (virtual screening) or designed de novo are highly desirable in the post-genomic area, many technical problems remain to be adequately addressed. This article presents an overview and discusses the limits of current computational methods for predicting the correct binding pose and accurate binding affinity. It also presents the performances of the most popular algorithms for exploring binary and multi-body protein interactions.

## 2011

• Transferable scoring function based on semiempirical quantum mechanical PM6-DH2 method: CDK2 with 15 structurally diverse inhibitors.
Dobes, Petr and Fanfrlík, Jindrich and Rezác, Jan and Otyepka, Michal and Hobza, Pavel
Journal of computer-aided molecular design, 2011, 25(3), 223-235
PMID: 21286784     doi: 10.1007/s10822-011-9413-5

A semiempirical quantum mechanical PM6-DH2 method accurately covering the dispersion interaction and H-bonding was used to score fifteen structurally diverse CDK2 inhibitors. The geometries of all the complexes were taken from the X-ray structures and were reoptimised by the PM6-DH2 method in continuum water. The total scoring function was constructed as an estimate of the binding free energy, i.e., as a sum of the interaction enthalpy, interaction entropy and the corrections for the inhibitor desolvation and deformation energies. The applied scoring function contains a clear thermodynamical terms and does not involve any adjustable empirical parameter. The best correlations with the experimental inhibition constants (ln K (i)) were found for bare interaction enthalpy (r (2)

• Assessing the performance of the molecular mechanics/Poisson Boltzmann surface area and molecular mechanics/generalized Born surface area methods. II. The accuracy of ranking poses generated from docking.
Hou, Tingjun and Wang, Junmei and Li, Youyong and Wang, Wei
Journal of computational chemistry, 2011, 32(5), 866-877
PMID: 20949517     doi: 10.1002/jcc.21666

In molecular docking, it is challenging to develop a scoring function that is accurate to conduct high-throughput screenings. Most scoring functions implemented in popular docking software packages were developed with many approximations for computational efficiency, which sacrifices the accuracy of prediction. With advanced technology and powerful computational hardware nowadays, it is feasible to use rigorous scoring functions, such as molecular mechanics/Poisson Boltzmann surface area (MM/PBSA) and molecular mechanics/generalized Born surface area (MM/GBSA) in molecular docking studies. Here, we systematically investigated the performance of MM/PBSA and MM/GBSA to identify the correct binding conformations and predict the binding free energies for 98 protein-ligand complexes. Comparison studies showed that MM/GBSA (69.4%) outperformed MM/PBSA (45.5%) and many popular scoring functions to identify the correct binding conformations. Moreover, we found that molecular dynamics simulations are necessary for some systems to identify the correct binding conformations. Based on our results, we proposed the guideline for MM/GBSA to predict the binding conformations. We then tested the performance of MM/GBSA and MM/PBSA to reproduce the binding free energies of the 98 protein-ligand complexes. The best prediction of MM/GBSA model with internal dielectric constant 2.0, produced a Spearman's correlation coefficient of 0.66, which is better than MM/PBSA (0.49) and almost all scoring functions used in molecular docking. In summary, MM/GBSA performs well for both binding pose predictions and binding free-energy estimations and is efficient to re-score the top-hit poses produced by other less-accurate scoring functions.

• Substantial improvements in large-scale redocking and screening using the novel HYDE scoring function.
Schneider, Nadine and Hindle, Sally and Lange, Gudrun and Klein, Robert and Albrecht, Jürgen and Briem, Hans and Beyer, Kristin and Clau{\ss}en, Holger and Gastreich, Marcus and Lemmen, Christian and Rarey, Matthias
Journal of computer-aided molecular design, 2011, 26(6), 701-723
PMID: 22203423     doi: 10.1007/s10822-011-9531-0

The HYDE scoring function consistently describes hydrogen bonding, the hydrophobic effect and desolvation. It relies on HYdration and DEsolvation terms which are calibrated using octanol/water partition coefficients of small molecules. We do not use affinity data for calibration, therefore HYDE is generally applicable to all protein targets. HYDE reflects the Gibbs free energy of binding while only considering the essential interactions of protein-ligand complexes. The greatest benefit of HYDE is that it yields a very intuitive atom-based score, which can be mapped onto the ligand and protein atoms. This allows the direct visualization of the score and consequently facilitates analysis of protein-ligand complexes during the lead optimization process. In this study, we validated our new scoring function by applying it in large-scale docking experiments. We could successfully predict the correct binding mode in 93% of complexes in redocking calculations on the Astex diverse set, while our performance in virtual screening experiments using the DUD dataset showed significant enrichment values with a mean AUC of 0.77 across all protein targets with little or no structural defects. As part of these studies, we also carried out a very detailed analysis of the data that revealed interesting pitfalls, which we highlight here and which should be addressed in future benchmark datasets.

• Correction to "A Machine Learning-Based Method To Improve Docking Scoring Functions and Its Application to Drug Repurposing"
Kinnings, Sarah L and Liu, Nina and Tonge, Peter J and Jackson, Richard M and Xie, Lei and Bourne, Philip E
Journal of chemical information and modeling, 2011, 51(5), 1195-1197
PMID: 21526828     doi: 10.1021/ci2001346

• MiniMuDS: a new optimizer using knowledge-based potentials improves scoring of docking solutions.
Spitzmüller, Andreas and Velec, Hans F G and Klebe, Gerhard
Journal of chemical information and modeling, 2011, 51(6), 1423-1430
PMID: 21528908     doi: 10.1021/ci200098v

In small molecule docking, the scoring and ranking of generated conformations is an important, though still not a completely resolved problem. Rescoring schemes often improve the quality of the obtained rankings. It is known that a local optimization is essential before a valid rescore value can be calculated. Here, we present a method that improves rescoring results obtained with the DrugScore function due to a new optimization technique. The method implements a more sophisticated search algorithm compared to the classic local optimization procedures used in this context. We validated the proposed method on a set of 192 protein-ligand complexes. Results show substantial improvements compared to original docking results with success rates increased by up to 10% for top scored solutions below 2\AA} root-mean-square deviation to the native state and up to 18% increase below 1\AA respectively.

• Normalizing Molecular Docking Rankings using Virtually Generated Decoys.
Wallach, Izhar and Jaitly, Navdeep and Nguyen, Kong and Schapira, Matthieu and Lilien, Ryan
Journal of chemical information and modeling, 2011, 51(8), 1817-1830
PMID: 21699246     doi: 10.1021/ci200175h

Drug discovery research often relies on the use of virtual screening via molecular docking to identify active hits in compound libraries. An area for improvement among many state-of-the-art docking methods is the accuracy of the scoring functions used to differentiate active from nonactive ligands. Many contemporary scoring functions are influenced by the physical properties of the docked molecule. This bias can cause molecules with certain physical properties to incorrectly score better than others. Since variation in physical properties is inevitable in large screening libraries, it is desirable to account for this bias. In this paper, we present a method of normalizing docking scores using virtually generated decoy sets with matched physical properties. First, our method generates a set of property-matched decoys for every molecule in the screening library. Each library molecule and its decoy set are docked using a state-of-the-art method, producing a set of raw docking scores. Next, the raw docking score of each library molecule is normalized against the scores of its decoys. The normalized score represents the probability that the raw docking score was drawn from the background distribution of nonactive property-matched decoys. Assuming that the distribution of scores of active molecules differs from the nonactive score distribution, we expect that the score of an active compound will have a low probability of having been drawn from the nonactive score distribution. In addition to the use of decoys in normalizing docking scores, we suggest that decoy sets may be a useful tool to evaluate, improve, or develop scoring functions. We show that by analyzing docking scores of library molecules with respect to the docking scores of their virtually generated property-matched decoys, one can gain insight into the advantages, limitations, and reliability of scoring functions.

• DEKOIS: Demanding Evaluation Kits for Objective in Silico Screening - A Versatile Tool for Benchmarking Docking Programs and Scoring Functions.
Vogel, Simon M and Bauer, Matthias R and Boeckler, Frank M
Journal of chemical information and modeling, 2011, 51(10), 2650-2665
PMID: 21774552     doi: 10.1021/ci2001549

For widely applied in silico screening techniques success depends on the rational selection of an appropriate method. We herein present a fast, versatile, and robust method to construct demanding evaluation kits for objective in silico screening (DEKOIS). This automated process enables creating tailor-made decoy sets for any given sets of bioactives. It facilitates a target-dependent validation of docking algorithms and scoring functions helping to save time and resources. We have developed metrics for assessing and improving decoy set quality and employ them to investigate how decoy embedding affects docking. We demonstrate that screening performance is target-dependent and can be impaired by latent actives in the decoy set (LADS) or enhanced by poor decoy embedding. The presented method allows extending and complementing the collection of publicly available high quality decoy sets toward new target space. All present and future DEKOIS data sets will be made accessible at www.dekois.com .

• Estimating binding affinities by docking/scoring methods using variable protonation states.
Park, Min-Sun and Gao, Cen and Stern, Harry A
Proteins, 2011, 79(1), 304-314
PMID: 21058298     doi: 10.1002/prot.22883

To investigate the effects of multiple protonation states on protein-ligand recognition, we generated alternative protonation states for selected titratable groups of ligands and receptors. The selection of states was based on the predicted pK(a) of the unbound receptor and ligand and the proximity of titratable groups of the receptor to the binding site. Various ligand tautomer states were also considered. An independent docking calculation was run for each state. Several protocols were examined: using an ensemble of all generated states of ligand and receptor, using only the most probable state of the unbound ligand/receptor, and using only the state giving the most favorable docking score. The accuracies of these approaches were compared, using a set of 176 protein-ligand complexes (15 receptors) for which crystal structures and measured binding affinities are available. The best agreement with experiment was obtained when ligand poses from experimental crystal structures were used. For 9 of 15 receptors, using an ensemble of all generated protonation states of the ligand and receptor gave the best correlation between calculated and measured affinities.

• Evaluation of Several Two-Step Scoring Functions Based on Linear Interaction Energy, Effective Ligand Size, and Empirical Pair Potentials for Prediction of Protein-Ligand Binding Geometry and Free Energy.
Rahaman, Obaidur and Estrada, Trilce P and Doren, Douglas J and Taufer, Michela and Brooks, Charles L and Armen, Roger S
Journal of chemical information and modeling, 2011, 51(9), 2047-2065
PMID: 21644546     doi: 10.1021/ci1003009

The performances of several two-step scoring approaches for molecular docking were assessed for their ability to predict binding geometries and free energies. Two new scoring functions designed for "step 2 discrimination" were proposed and compared to our CHARMM implementation of the linear interaction energy (LIE) approach using the Generalized-Born with Molecular Volume (GBMV) implicit solvation model. A scoring function S1 was proposed by considering only "interacting" ligand atoms as the "effective size" of the ligand and extended to an empirical regression-based pair potential S2. The S1 and S2 scoring schemes were trained and 5-fold cross-validated on a diverse set of 259 protein-ligand complexes from the Ligand Protein Database (LPDB). The regression-based parameters for S1 and S2 also demonstrated reasonable transferability in the CSARdock 2010 benchmark using a new data set (NRC HiQ) of diverse protein-ligand complexes. The ability of the scoring functions to accurately predict ligand geometry was evaluated by calculating the discriminative power (DP) of the scoring functions to identify native poses. The parameters for the LIE scoring function with the optimal discriminative power (DP) for geometry (step 1 discrimination) were found to be very similar to the best-fit parameters for binding free energy over a large number of protein-ligand complexes (step 2 discrimination). Reasonable performance of the scoring functions in enrichment of active compounds in four different protein target classes established that the parameters for S1 and S2 provided reasonable accuracy and transferability. Additional analysis was performed to definitively separate scoring function performance from molecular weight effects. This analysis included the prediction of ligand binding efficiencies for a subset of the CSARdock NRC HiQ data set where the number of ligand heavy atoms ranged from 17 to 35. This range of ligand heavy atoms is where improved accuracy of predicted ligand efficiencies is most relevant to real-world drug design efforts.

• PLS-DA - Docking Optimized Combined Energetic Terms (PLSDA-DOCET) protocol: a brief evaluation.
Avram, Sorin and Pacureanu, Liliana Mioara and Seclaman, Edward and Bora, Alina and Kurunczi, Ludovic G
Journal of chemical information and modeling, 2011, 51(12), 3169-3179
PMID: 22066983     doi: 10.1021/ci2002268

Docking studies have become popular approaches in drug design, where the binding energy of the ligand in the active site of the protein is estimated by a scoring function. Many promising techniques were developed to enhance the performance of scoring functions including the fusion of multiple scoring functions outcomes into a so-called consensus scoring function. Hereby, we evaluated the target oriented consensus technique using the energetic terms of several scoring functions. The approach was denoted PLSDA-DOCET. Optimization strategies for consensus energetic terms and scoring functions based on ROC metric were compared to classical rigid docking and to ligand-based similarity search methods comprising 2D fingerprints and ROCS. The ROCS results indicate large performance variations depending on the biological target. The AUC-based strategy of PLSDA-DOCET outperformed the other docking approaches regarding simple retrieval and scaffold-hopping. The superior performance of PLSDA-DOCET protocol relative to single and combined scoring functions was validated on an external test set. We found a relative low mean correlation of the ranks of the chemotypes retrieved by the PLSDA-DOCET protocol and all the other methods employed here.

• Statistical potential for modeling and ranking of protein-ligand interactions.
Fan, Hao and Schneidman-Duhovny, Dina and Irwin, John J and Dong, Guangqiang and Shoichet, Brian K and Sali, Andrej
Journal of chemical information and modeling, 2011, 51(12), 3078-3092
PMID: 22014038     doi: 10.1021/ci200377u

Applications in structural biology and medicinal chemistry require protein-ligand scoring functions for two distinct tasks: (i) ranking different poses of a small molecule in a protein binding site and (ii) ranking different small molecules by their complementarity to a protein site. Using probability theory, we developed two atomic distance-dependent statistical scoring functions: PoseScore was optimized for recognizing native binding geometries of ligands from other poses and RankScore was optimized for distinguishing ligands from nonbinding molecules. Both scores are based on a set of 8,885 crystallographic structures of protein-ligand complexes but differ in the values of three key parameters. Factors influencing the accuracy of scoring were investigated, including the maximal atomic distance and non-native ligand geometries used for scoring, as well as the use of protein models instead of crystallographic structures for training and testing the scoring function. For the test set of 19 targets, RankScore improved the ligand enrichment (logAUC) and early enrichment (EF(1)) scores computed by DOCK 3.6 for 13 and 14 targets, respectively. In addition, RankScore performed better at rescoring than each of seven other scoring functions tested. Accepting both the crystal structure and decoy geometries with all-atom root-mean-square errors of up to 2\AA} from the crystal structure as correct binding poses, PoseScore gave the best score to a correct binding pose among 100 decoys for 88% of all cases in a benchmark set containing 100 protein-ligand complexes. PoseScore accuracy is comparable to that of DrugScore(CSD) and ITScore/SE and superior to 12 other tested scoring functions. Therefore, RankScore can facilitate ligand discovery, by ranking complexes of the target with different small molecules; PoseScore can be used for protein-ligand complex structure prediction, by ranking different conformations of a given protein-ligand pair. The statistical potentials are available through the Integrative Modeling Platform (IMP) software package (http://salilab.org/imp) and the LigScore Web server (http://salilab.org/ligscore/).

• Improving molecular docking through eHiTS' tunable scoring function.
Ravitz, Orr and Zsoldos, Zsolt and Simon, Aniko
Journal of computer-aided molecular design, 2011, 25(11), 1033-1051
PMID: 22076470     doi: 10.1007/s10822-011-9482-5

We present three complementary approaches for score-tuning that improve docking performance in pose prediction, virtual screening and binding affinity assessment. The methodology utilizes experimental data to customize the scoring function for the system of interest considering the specific docking scenario. The tuning approach, which has been implemented as an automated utility in eHiTS, is introduced as a solution to one of the conundrums of the molecular docking paradigm, namely, the lack of a universally well performing scoring function. The accuracy of scoring functions has been shown to be generally system-dependent, and particularly lacking for binding energy and bio-activity predictions. In the proposed approach, pose and energy predictions are enhanced by adjusting the relative weights of the eHiTS energy terms to improve score-RMSD or score-affinity correlations. In a virtual screening context ligand-based similarity is used to rescale the docking score such that better enrichment factors are achieved. We discuss the algorithmic details of the methods, and demonstrate the effects of score tuning on a variety of targets, including CDK2, BACE1 and neuraminidase, as well as on the popular benchmarks-the Directory of Useful Decoys and the PDBBind database.

• BEAR, a novel virtual screening methodology for drug discovery.
Degliesposti, Gianluca and Portioli, Corinne and Parenti, Marco Daniele and Rastelli, Giulio
Journal of biomolecular screening, 2011, 16(1), 129-133
PMID: 21084717     doi: 10.1177/1087057110388276

BEAR (binding estimation after refinement) is a new virtual screening technology based on the conformational refinement of docking poses through molecular dynamics and prediction of binding free energies using accurate scoring functions. Here, the authors report the results of an extensive benchmark of the BEAR performance in identifying a smaller subset of known inhibitors seeded in a large (1.5 million) database of compounds. BEAR performance proved strikingly better if compared with standard docking screening methods. The validations performed so far showed that BEAR is a reliable tool for drug discovery. It is fast, modular, and automated, and it can be applied to virtual screenings against any biological target with known structure and any database of compounds.

• Knowledge-Based Scoring Functions in Drug Design: 3. A Two-Dimensional Knowledge-Based Hydrogen-Bonding Potential for the Prediction of Protein-Ligand Interactions.
Zheng, Mingyue and Xiong, Bing and Luo, Cheng and Li, Shanshan and Liu, Xian and Shen, Qianchen and Li, Jing and Zhu, Weiliang and Luo, Xiaomin and Jiang, Hualiang
Journal of chemical information and modeling, 2011, 50(11), 2994-3004
PMID: 21999432     doi: 10.1021/ci2003939

Hydrogen bonding is a key contributor to the molecular recognition between ligands and their host molecules in biological systems. Here we develop a novel orientation-dependent hydrogen bonding potential based on the geometric characteristics of hydrogen bonds observed in 44,585 protein-ligand complexes. We find a close correspondence between the empirical knowledge and the energy landscape inferred from the distribution of HBs. A scoring function based on the resultant hydrogen-bonding potentials discriminates native protein-ligand structures from incorrectly docked decoys with remarkable predictive power.

• A machine learning-based method to improve docking scoring functions and its application to drug repurposing.
Kinnings, Sarah L and Liu, Nina and Tonge, Peter J and Jackson, Richard M and Xie, Lei and Bourne, Philip E
Journal of chemical information and modeling, 2011, 51(2), 408-419
PMID: 21291174     doi: 10.1021/ci100369f

Docking scoring functions are notoriously weak predictors of binding affinity. They typically assign a common set of weights to the individual energy terms that contribute to the overall energy score; however, these weights should be gene family dependent. In addition, they incorrectly assume that individual interactions contribute toward the total binding affinity in an additive manner. In reality, noncovalent interactions often depend on one another in a nonlinear manner. In this paper, we show how the use of support vector machines (SVMs), trained by associating sets of individual energy terms retrieved from molecular docking with the known binding affinity of each compound from high-throughput screening experiments, can be used to improve the correlation between known binding affinities and those predicted by the docking program eHiTS. We construct two prediction models: a regression model trained using IC(50) values from BindingDB, and a classification model trained using active and decoy compounds from the Directory of Useful Decoys (DUD). Moreover, to address the issue of overrepresentation of negative data in high-throughput screening data sets, we have designed a multiple-planar SVM training procedure for the classification model. The increased performance that both SVMs give when compared with the original eHiTS scoring function highlights the potential for using nonlinear methods when deriving overall energy scores from their individual components. We apply the above methodology to train a new scoring function for direct inhibitors of Mycobacterium tuberculosis (M.tb) InhA. By combining ligand binding site comparison with the new scoring function, we propose that phosphodiesterase inhibitors can potentially be repurposed to target M.tb InhA. Our methodology may be applied to other gene families for which target structures and activity data are available, as demonstrated in the work presented here.

• AADS - An Automated Active Site Identification, Docking, and Scoring Protocol for Protein Targets Based on Physicochemical Descriptors.
Singh, Tanya and Biswas, D and Jayaram, B.
Journal of chemical information and modeling, 2011, 51(10), 2515-2527
PMID: 21877713     doi: 10.1021/ci200193z

We report here a robust automated active site detection, docking, and scoring (AADS) protocol for proteins with known structures. The active site finder identifies all cavities in a protein and scores them based on the physicochemical properties of functional groups lining the cavities in the protein. The accuracy realized on 620 proteins with sizes ranging from 100 to 600 amino acids with known drug active sites is 100% when the top ten cavity points are considered. These top ten cavity points identified are then submitted for an automated docking of an input ligand/candidate molecule. The docking protocol uses an all atom energy based Monte Carlo method. Eight low energy docked structures corresponding to different locations and orientations of the candidate molecule are stored at each cavity point giving 80 docked structures overall which are then ranked using an effective free energy function and top five structures are selected. The predicted structure and energetics of the complexes agree quite well with experiment when tested on a data set of 170 protein-ligand complexes with known structures and binding affinities. The AADS methodology is implemented on an 80 processor cluster and presented as a freely accessible, easy to use tool at http://www.scfbio-iitd.res.in/dock/ActiveSite_new.jsp .

• Virtual screening using molecular simulations.
Yang, Tianyi and Wu, Johnny C and Yan, Chunli and Wang, Yuanfeng and Luo, Ray and Gonzales, Michael B and Dalby, Kevin N and Ren, Pengyu
Proteins, 2011, 79(6), 1940-1951
PMID: 21491494     doi: 10.1002/prot.23018

Effective virtual screening relies on our ability to make accurate prediction of protein-ligand binding, which remains a great challenge. In this work, utilizing the molecular-mechanics Poisson-Boltzmann (or Generalized Born) surface area approach, we have evaluated the binding affinity of a set of 156 ligands to seven families of proteins, trypsin $\beta$, thrombin $\alpha$, cyclin-dependent kinase (CDK), cAMP-dependent kinase (PKA), urokinase-type plasminogen activator, $\beta$-glucosidase A, and coagulation factor Xa. The effect of protein dielectric constant in the implicit-solvent model on the binding free energy calculation is shown to be important. The statistical correlations between the binding energy calculated from the implicit-solvent approach and experimental free energy are in the range of 0.56-0.79 across all the families. This performance is better than that of typical docking programs especially given that the latter is directly trained using known binding data whereas the molecular mechanics is based on general physical parameters. Estimation of entropic contribution remains the barrier to accurate free energy calculation. We show that the traditional rigid rotor harmonic oscillator approximation is unable to improve the binding free energy prediction. Inclusion of conformational restriction seems to be promising but requires further investigation. On the other hand, our preliminary study suggests that implicit-solvent based alchemical perturbation, which offers explicit sampling of configuration entropy, can be a viable approach to significantly improve the prediction of binding free energy. Overall, the molecular mechanics approach has the potential for medium to high-throughput computational drug discovery.

• SwissParam: a fast force field generation tool for small organic molecules.
Zoete, Vincent and Cuendet, Michel A and Grosdidier, Aurélien and Michielin, Olivier
Journal of computational chemistry, 2011, 32(11), 2359-2368
PMID: 21541964     doi: 10.1002/jcc.21816

The drug discovery process has been deeply transformed recently by the use of computational ligand-based or structure-based methods, helping the lead compounds identification and optimization, and finally the delivery of new drug candidates more quickly and at lower cost. Structure-based computational methods for drug discovery mainly involve ligand-protein docking and rapid binding free energy estimation, both of which require force field parameterization for many drug candidates. Here, we present a fast force field generation tool, called SwissParam, able to generate, for arbitrary small organic molecule, topologies, and parameters based on the Merck molecular force field, but in a functional form that is compatible with the CHARMM force field. Output files can be used with CHARMM or GROMACS. The topologies and parameters generated by SwissParam are used by the docking software EADock2 and EADock DSS to describe the small molecules to be docked, whereas the protein is described by the CHARMM force field, and allow them to reach success rates ranging from 56 to 78%. We have also developed a rapid binding free energy estimation approach, using SwissParam for ligands and CHARMM22/27 for proteins, which requires only a short minimization to reproduce the experimental binding free energy of 214 ligand-protein complexes involving 62 different proteins, with a standard error of 2.0 kcal mol(-1), and a correlation coefficient of 0.74. Together, these results demonstrate the relevance of using SwissParam topologies and parameters to describe small organic molecules in computer-aided drug design applications, together with a CHARMM22/27 description of the target protein. SwissParam is available free of charge for academic users at www.swissparam.ch.

• Computer-aided drug design platform using PyMOL.
Lill, Markus A and Danielson, Matthew L
Journal of computer-aided molecular design, 2011, 25(1), 13-19
PMID: 21053052     doi: 10.1007/s10822-010-9395-8

The understanding and optimization of protein-ligand interactions are instrumental to medicinal chemists investigating potential drug candidates. Over the past couple of decades, many powerful standalone tools for computer-aided drug discovery have been developed in academia providing insight into protein-ligand interactions. As programs are developed by various research groups, a consistent user-friendly graphical working environment combining computational techniques such as docking, scoring, molecular dynamics simulations, and free energy calculations is needed. Utilizing PyMOL we have developed such a graphical user interface incorporating individual academic packages designed for protein preparation (AMBER package and Reduce), molecular mechanics applications (AMBER package), and docking and scoring (AutoDock Vina and SLIDE). In addition to amassing several computational tools under one interface, the computational platform also provides a user-friendly combination of different programs. For example, utilizing a molecular dynamics (MD) simulation performed with AMBER as input for ensemble docking with AutoDock Vina. The overarching goal of this work was to provide a computational platform that facilitates medicinal chemists, many who are not experts in computational methodologies, to utilize several common computational techniques germane to drug discovery. Furthermore, our software is open source and is aimed to initiate collaborative efforts among computational researchers to combine other open source computational methods under a single, easily understandable graphical user interface.

• Can We Trust Docking Results? Evaluation of Seven Commonly Used Programs on PDBbind Database
Plewczynski, Dariusz and Lazniewski, Michal and Augustyniak, Rafal and Ginalski, Krzysztof
Journal of computational chemistry, 2011, 32(4), 742-755
PMID: 20812323     doi: 10.1002/jcc.21643

Docking is one of the most commonly used techniques in drug design. It is used for both identifying correct poses of a ligand in the binding site of a protein as well as for the estimation of the strength of protein-ligand interaction. Because millions of compounds must be screened, before a suitable target for biological testing can be identified, all calculations should be done in a reasonable time frame. Thus, all programs currently in use exploit empirically based algorithms, avoiding systematic search of the conformational space. Similarly, the scoring is done using simple equations, which makes it possible to speed up the entire process. Therefore, docking results have to be verified by subsequent in vitro studies. The purpose of our work was to evaluate seven popular docking programs (Surf lex, LigandFit, Glide, GOLD, FlexX, eHiTS, and Auto Dock) on the extensive dataset composed of 1300 protein-ligands complexes from PDBbind 2007 database, where experimentally measured binding affinity values were also available. We compared independently the ability of proper posing [according to Root mean square deviation (or Root mean square distance) of predicted conformations versus the corresponding native one] and scoring (by calculating the correlation between docking score and ligand binding strength). To our knowledge, it is the first large-scale docking evaluation that covers both aspects of docking programs, that is, predicting ligand conformation and calculating the strength of its binding. More than 1000 protein-ligand pairs cover a wide range of different protein families and inhibitor classes. Our results clearly showed that the ligand binding conformation could be identified in most cases by using the existing software, yet we still observed the lack of universal scoring function for all types of molecules and protein families. (C) 2010 Wiley Periodicals, Inc. J Comput Chem 32: 742-755, 2011

• NNScore 2.0: a neural-network receptor-ligand scoring function.
Durrant, Jacob D and McCammon, J Andrew
Journal of chemical information and modeling, 2011, 51(11), 2897-2903
PMID: 22017367     doi: 10.1021/ci2003889

NNScore is a neural-network-based scoring function designed to aid the computational identification of small-molecule ligands. While the test cases included in the original NNScore article demonstrated the utility of the program, the application examples were limited. The purpose of the current work is to further confirm that neural-network scoring functions are effective, even when compared to the scoring functions of state-of-the-art docking programs, such as AutoDock, the most commonly cited program, and AutoDock Vina, thought to be two orders of magnitude faster. Aside from providing additional validation of the original NNScore function, we here present a second neural-network scoring function, NNScore 2.0. NNScore 2.0 considers many more binding characteristics when predicting affinity than does the original NNScore. The network output of NNScore 2.0 also differs from that of NNScore 1.0; rather than a binary classification of ligand potency, NNScore 2.0 provides a single estimate of the pK(d). To facilitate use, NNScore 2.0 has been implemented as an open-source python script. A copy can be obtained from http://www.nbcr.net/software/nnscore/ .

• Implementation and evaluation of a docking-rescoring method using molecular footprint comparisons.
Balius, Trent E and Mukherjee, Sudipto and Rizzo, Robert C
Journal of computational chemistry, 2011, 32(10), 2273-2289
PMID: 21541962     doi: 10.1002/jcc.21814

A docking-rescoring method, based on per-residue van der Waals (VDW), electrostatic (ES), or hydrogen bond (HB) energies has been developed to aid discovery of ligands that have interaction signatures with a target (footprints) similar to that of a reference. Biologically useful references could include known drugs, inhibitors, substrates, transition states, or side-chains that mediate protein-protein interactions. Termed footprint similarity (FPS) score, the method, as implemented in the program DOCK, was validated and characterized using: (1) pose identification, (2) crossdocking, (3) enrichment, and (4) virtual screening. Improvements in pose identification (6-12%) were obtained using footprint-based (FPSVDW+ES) vs. standard DOCK (DCEVDW+ES) scoring as evaluated on three large datasets (680-775 systems) from the SB2010 database. Enhanced pose identification was also observed using FPS (45.4% or 70.9%) compared with DCE (17.8%) methods to rank challenging crossdocking ensembles from carbonic anhydrase. Enrichment tests, for three representative systems, revealed FPSVDW+ES scoring yields significant early fold enrichment in the top 10% of ranked databases. For EGFR, top FPS poses are nicely accommodated in the molecular envelope defined by the reference in comparison with DCE, which yields distinct molecular weight bias toward larger molecules. Results from a representative virtual screen of ca. 1 million compounds additionally illustrate how ligands with footprints similar to a known inhibitor can readily be identified from within large commercially available databases. By providing an alternative way to rank ligand poses in a simple yet directed manner we anticipate that FPS scoring will be a useful tool for docking and structure-based design. (C) 2011 Wiley Periodicals, Inc. J Comput Chem 32: 2273-2289, 2011

• DSX: a knowledge-based scoring function for the assessment of protein-ligand complexes.
Neudert, Gerd and Klebe, Gerhard
Journal of chemical information and modeling, 2011, 51(10), 2731-2745
PMID: 21863864     doi: 10.1021/ci200274q

We introduce the new knowledge-based scoring function DSX that consists of distance-dependent pair potentials, novel torsion angle potentials, and newly defined solvent accessible surface-dependent potentials. DSX pair potentials are based on the statistical formalism of DrugScore, extended by a much more specialized set of atom types. The original DrugScore-like reference state is rather unstable with respect to modifications in the used atom types. Therefore, an important method to overcome this problem and to allow for robust results when deriving pair potentials for arbitrary sets of atom types is presented. A validation based on a carefully prepared test set is shown, enabling direct comparison to the majority of other popular scoring functions. Here, DSX features superior performance with respect to docking- and ranking power and runtime requirements. Furthermore, the beneficial combination with torsion angle-dependent and desolvation-dependent potentials is demonstrated. DSX is robust, flexible, and capable of working together with special features of popular docking engines, e.g., flexible protein residues in AutoDock or GOLD. The program is freely available to the scientific community and can be downloaded from our Web site www.agklebe.de .

• Exhaustive search and solvated interaction energy (SIE) for virtual screening and affinity prediction.
Sulea, Traian and Hogues, Hervé and Purisima, Enrico O
Journal of computer-aided molecular design, 2011, 26(5), 617-633
PMID: 22198519     doi: 10.1007/s10822-011-9529-7

We carried out a prospective evaluation of the utility of the SIE (solvation interaction energy) scoring function for virtual screening and binding affinity prediction. Since experimental structures of the complexes were not provided, this was an exercise in virtual docking as well. We used our exhaustive docking program, Wilma, to provide high-quality poses that were rescored using SIE to provide binding affinity predictions. We also tested the combination of SIE with our latest solvation model, first shell of hydration (FiSH), which captures some of the discrete properties of water within a continuum model. We achieved good enrichment in virtual screening of fragments against trypsin, with an area under the curve of about 0.7 for the receiver operating characteristic curve. Moreover, the early enrichment performance was quite good with 50% of true actives recovered with a 15% false positive rate in a prospective calculation and with a 3% false positive rate in a retrospective application of SIE with FiSH. Binding affinity predictions for both trypsin and host-guest complexes were generally within 2 kcal/mol of the experimental values. However, the rank ordering of affinities differing by 2 kcal/mol or less was not well predicted. On the other hand, it was encouraging that the incorporation of a more sophisticated solvation model into SIE resulted in better discrimination of true binders from binders. This suggests that the inclusion of proper Physics in our models is a fruitful strategy for improving the reliability of our binding affinity predictions.

• Comments on "leave-cluster-out cross-validation is appropriate for scoring functions derived from diverse protein data sets": significance for the validation of scoring functions.
Ballester, Pedro J and Mitchell, John B O
Journal of chemical information and modeling, 2011, 51(8), 1739-1741
PMID: 21591735     doi: 10.1021/ci200057e

• Predicting Fragment Binding Poses Using a Combined MCSS MM-GBSA Approach.
Haider, Muhammad K and Bertrand, Hugues-Olivier and Hubbard, Roderick E
Journal of chemical information and modeling, 2011, 51(5), 1092-1105
PMID: 21528911     doi: 10.1021/ci100469n

Improved methods are required to predict the position and orientation (pose) of binding to the target protein of low molecular weight compounds identified in fragment screening campaigns. This is particularly important to guide initial chemistry to generate structure-activity relationships for the cases where a high resolution structure cannot be obtained. We have assessed the benefit of an implicit solvent method for assessment of fragment binding poses generated by the Multiple Copy Simultaneous Search (MCSS) method in CHARMm. Additionally, the effect of using multiple receptor structures for a flexible receptor is investigated. The original MCSS performance -50% of fragment positions accurately predicted and scored - was increased up to 67% by scoring MCSS energy minima with a Molecular Mechanics Generalized Born approach with molecular volume integration and Surface Area model (MM-GBSA). The same increase in performance (but occasionally for different targets) was observed when using the docking program GOLD followed by MM-GBSA rescoring. The combined results from both methods resulted in a higher success rate emphasizing that a comparison of different docking methods can increase the correct identification of binding poses. For a receptor where multiple structures are available, Hsp90, the average performance on randomly adding receptor structures was also investigated. The results suggest that predictions using these docking methods can be used with some confidence to guide chemical optimization, if the structure of the target either remains relatively fixed on ligand binding, or if a number of crystal structures are available with diverse ligands bound and there is information on the positions of key water molecules in the binding site.

## 2010

• Chemical space sampling by different scoring functions and crystal structures.
Brooijmans, Natasja and Humblet, Christine
Journal of computer-aided molecular design, 2010, 24(5), 433-447
PMID: 20401681     doi: 10.1007/s10822-010-9356-2

Virtual screening has become a popular tool to identify novel leads in the early phases of drug discovery. A variety of docking and scoring methods used in virtual screening have been the subject of active research in an effort to gauge limitations and articulate best practices. However, how to best utilize different scoring functions and various crystal structures, when available, is not yet well understood. In this work we use multiple crystal structures of PI3 K-gamma in both prospective and retrospective virtual screening experiments. Both Glide SP scoring and Prime MM-GBSA rescoring are utilized in the prospective and retrospective virtual screens, and consensus scoring is investigated in the retrospective virtual screening experiments. The results show that each of the different crystal structures that was used, samples a different chemical space, i.e. different chemotypes are prioritized by each structure. In addition, the different (re)scoring functions prioritize different chemotypes as well. Somewhat surprisingly, the Prime MM-GBSA scoring function generally gives lower enrichments than Glide SP. Finally we investigate the impact of different ligand preparation protocols on virtual screening enrichment factors. In summary, different crystal structures and different scoring functions are complementary to each other and allow for a wider variety of chemotypes to be considered for experimental follow-up.

• Prediction of protein-ligand binding affinities using multiple instance learning.
Teramoto, Reiji and Kashima, Hisashi
Journal of molecular graphics & modelling, 2010, 29(3), 492-497
PMID: 20965757     doi: 10.1016/j.jmgm.2010.09.006

Accurate prediction of protein-ligand binding affinities for lead optimization in drug discovery remains an important and challenging problem on scoring functions for docking simulation. In this paper, we propose a data-driven approach that integrates multiple scoring functions to predict protein-ligand binding affinity directly. We then propose a new method called multiple instance regression based scoring (MIRS) that incorporates unbound ligand conformations using multiple scoring functions. We evaluated the predictive performance of MIRS using 100 protein-ligand complexes and their binding affinities. The experimental results showed that MIRS outperformed the 11 conventional scoring functions including LigScore, PLP, AutoDock, G-Score, D-Score, LUDI, F-Score, ChemScore, X-Score, PMF, and DrugScore. In addition, we confirmed that MIRS performed well on binding pose prediction. Our results reveal that it is indispensable to incorporate unbound ligand conformations in both binding affinity prediction and binding pose prediction. The proposed method will accelerate efficient lead optimization on structure-based drug design and provide a new direction to designing of new scoring score functions.

• A fast protein-ligand docking algorithm based on hydrogen bond matching and surface shape complementarity.
Luo, Wenjia and Pei, Jianfeng and Zhu, Yushan
Journal of Molecular Modeling, 2010, 16(5), 903-913
PMID: 19823881     doi: 10.1007/s00894-009-0598-7

With the rapid development of structural determination of target proteins for human diseases, high throughout virtual screening based drug discovery is gaining popularity gradually. In this paper, a fast docking algorithm (H-DOCK) based on hydrogen bond matching and surface shape complementarity was developed. In H-DOCK, firstly a divide-and-conquer strategy based enumeration approach is applied to rank the intermolecular modes between protein and ligand by maximizing their hydrogen bonds matching, then each docked conformation of the ligand is calculated according to the matched hydrogen bonding geometry, finally a simple but effective scoring function reflecting mainly the van der Waals interaction is used to evaluate the docked conformations of the ligand. H-DOCK is tested for rigid ligand docking and flexible one, the latter is implemented by repeating rigid docking for multiple conformations of a small molecule and ranking all together. For rigid ligands, H-DOCK was tested on a set of 271 complexes where there is at least one intermolecular hydrogen bond, and H-DOCK achieved success rate (RMSD<2.0 A) of 91.1%. For flexible ligands, H-DOCK was tested on another set of 93 complexes, where each case was a conformation ensemble containing native ligand conformation as well as 100 decoy ones generated by AutoDock, and the success rate reached 81.7%. The high success rate of H-DOCK indicates that the hydrogen bonding and steric hindrance can grasp the key interaction between protein and ligand. H-DOCK is quite efficient compared with the conventional docking algorithms, and it takes only about 0.14 seconds for a rigid ligand docking and about 8.25 seconds for a flexible one on average. According to the preliminary docking results, it implies that H-DOCK can be potentially used for large scale virtual screening as a pre-filter for a more accurate but less efficient docking algorithm.

• Use of the FACTS solvation model for protein-ligand docking calculations. Application to EADock.
Zoete, Vincent and Grosdidier, Aurélien and Cuendet, Michel and Michielin, Olivier
Journal of molecular recognition : JMR, 2010, 23(5), 457-461
PMID: 20101644     doi: 10.1002/jmr.1012

Protein-ligand docking has made important progress during the last decade and has become a powerful tool for drug development, opening the way to virtual high throughput screening and in silico structure-based ligand design. Despite the flattering picture that has been drawn, recent publications have shown that the docking problem is far from being solved, and that more developments are still needed to achieve high successful prediction rates and accuracy. Introducing an accurate description of the solvation effect upon binding is thought to be essential to achieve this goal. In particular, EADock uses the Generalized Born Molecular Volume 2 (GBMV2) solvent model, which has been shown to reproduce accurately the desolvation energies calculated by solving the Poisson equation. Here, the implementation of the Fast Analytical Continuum Treatment of Solvation (FACTS) as an implicit solvation model in small molecules docking calculations has been assessed using the EADock docking program. Our results strongly support the use of FACTS for docking. The success rates of EADock/FACTS and EADock/GBMV2 are similar, i.e. around 75% for local docking and 65% for blind docking. However, these results come at a much lower computational cost: FACTS is 10 times faster than GBMV2 in calculating the total electrostatic energy, and allows a speed up of EADock by a factor of 4. This study also supports the EADock development strategy relying on the CHARMM package for energy calculations, which enables straightforward implementation and testing of the latest developments in the field of Molecular Modeling.

• SKATE: a docking program that decouples systematic sampling from scoring.
Feng, Jianwen A and Marshall, Garland R.
Journal of computational chemistry, 2010, 31(14), 2540-2554
PMID: 20740553     doi: 10.1002/jcc.21545

SKATE is a docking prototype that decouples systematic sampling from scoring. This novel approach removes any interdependence between sampling and scoring functions to achieve better sampling and, thus, improves docking accuracy. SKATE systematically samples a ligand's conformational, rotational and translational degrees of freedom, as constrained by a receptor pocket, to find sterically allowed poses. Efficient systematic sampling is achieved by pruning the combinatorial tree using aggregate assembly, discriminant analysis, adaptive sampling, radial sampling, and clustering. Because systematic sampling is decoupled from scoring, the poses generated by SKATE can be ranked by any published, or in-house, scoring function. To test the performance of SKATE, ligands from the Asetex/CDCC set, the Surflex set, and the Vertex set, a total of 266 complexes, were redocked to their respective receptors. The results show that SKATE was able to sample poses within 2 A RMSD of the native structure for 98, 95, and 98% of the cases in the Astex/CDCC, Surflex, and Vertex sets, respectively. Cross-docking accuracy of SKATE was also assessed by docking 10 ligands to thymidine kinase and 73 ligands to cyclin-dependent kinase.

• NNScore: a neural-network-based scoring function for the characterization of protein-ligand complexes.
Durrant, Jacob D and McCammon, J Andrew
Journal of chemical information and modeling, 2010, 50(10), 1865-1871
PMID: 20845954     doi: 10.1021/ci100244v

As high-throughput biochemical screens are both expensive and labor intensive, researchers in academia and industry are turning increasingly to virtual-screening methodologies. Virtual screening relies on scoring functions to quickly assess ligand potency. Although useful for in silico ligand identification, these scoring functions generally give many false positives and negatives; indeed, a properly trained human being can often assess ligand potency by visual inspection with greater accuracy. Given the success of the human mind at protein-ligand complex characterization, we present here a scoring function based on a neural network, a computational model that attempts to simulate, albeit inadequately, the microscopic organization of the brain. Computer-aided drug design depends on fast and accurate scoring functions to aid in the identification of small-molecule ligands. The scoring function presented here, used either on its own or in conjunction with other more traditional functions, could prove useful in future drug-discovery efforts.

• A new Lamarckian genetic algorithm for flexible ligand-receptor docking.
Fuhrmann, Jan and Rurainski, Alexander and Lenhof, Hans-Peter and Neumann, Dirk
Journal of computational chemistry, 2010, 31(9), 1911-1918
PMID: 20082382     doi: 10.1002/jcc.21478

We present a Lamarckian genetic algorithm (LGA) variant for flexible ligand-receptor docking which allows to handle a large number of degrees of freedom. Our hybrid method combines a multi-deme LGA with a recently published gradient-based method for local optimization of molecular complexes. We compared the performance of our new hybrid method to two non gradient-based search heuristics on the Astex diverse set for flexible ligand-receptor docking. Our results show that the novel approach is clearly superior to other LGAs employing a stochastic optimization method. The new algorithm features a shorter run time and gives substantially better results, especially with increasing complexity of the ligands. Thus, it may be used to dock ligands with many rotatable bonds with high efficiency.

• An interaction-motif-based scoring function for protein-ligand docking.
Xie, Zhong-Ru and Hwang, Ming-Jing
Bmc Bioinformatics, 2010, 11, 298
PMID: 20525216     doi: 10.1186/1471-2105-11-298

BACKGROUND:A good scoring function is essential for molecular docking computations. In conventional scoring functions, energy terms modeling pairwise interactions are cumulatively summed, and the best docking solution is selected. Here, we propose to transform protein-ligand interactions into three-dimensional geometric networks, from which recurring network substructures, or network motifs, are selected and used to provide probability-ranked interaction templates with which to score docking solutions.

## 2009

• Elastic potential grids: accurate and efficient representation of intermolecular interactions for fully flexible docking.
Kazemi, Sina and Krüger, Dennis M and Sirockin, Finton and Gohlke, Holger
Chemmedchem, 2009, 4(8), 1264-1268
PMID: 19514026     doi: 10.1002/cmdc.200900146

• Scoring confidence index: statistical evaluation of ligand binding mode predictions.
Zavodszky, Maria I and Stumpff-Kane, Andrew W and Lee, David J and Feig, Michael
Journal of computer-aided molecular design, 2009, 23(5), 289-299
PMID: 19153808     doi: 10.1007/s10822-008-9258-8

Protein-ligand docking programs can generate a large number of possible binding orientations for each ligand candidate. The challenge is to identify the orientations closest to the native binding mode using a scoring method. Many different scoring functions have been developed for protein-ligand scoring, but their performance on binding mode prediction is often target-dependent. In this study, a statistical approach was employed to provide a confidence measure of scoring performance in finding close to the correct docked ligand orientations. It exploits the fact that the scores provided by an adequately performing scoring function generally improve as the ligand binding modes get closer to the correct native orientation. For such cases, the correlation coefficient of scores versus distances is expected to be highest when the most native-like orientation is used as a reference. This correlation coefficient, called the correlation-based score (CBScore), was used as an indicator of how far the docked pose was from the native orientation. The correlation between the original scores and CBScores as well as the range of CBScores were found to be good measures of scoring performance. They were combined into a single quantity, called the scoring confidence index. High values of the scoring confidence index were indicative of pronounced and relatively smooth binding energy landscapes with easily discernable global minima, resulting in reliable binding mode predictions. Low values of this index reflected rugged energy landscapes making the prediction of the correct binding mode very difficult and often unreliable. The diagnostic ability of the scoring confidence index was tested on a non-redundant set of 50 protein-ligand complexes scored with three commonly employed scoring functions: AffiScore, DrugScore and X-Score. Binding mode predictions were found to be three times more reliable for complexes with scoring confidence indices in the upper half than for cases with values in the lower half of the resulting range of 0-1.6. This new confidence measure of scoring performance is expected to be a valuable tool for virtual screening applications.

• Scoring ligand similarity in structure-based virtual screening.
Zavodszky, Maria I and Rohatgi, Anjali and Van Voorst, Jeffrey R and Yan, Honggao and Kuhn, Leslie A
Journal of molecular recognition : JMR, 2009, 22(4), 280-292
PMID: 19235177     doi: 10.1002/jmr.942

Scoring to identify high-affinity compounds remains a challenge in virtual screening. On one hand, protein-ligand scoring focuses on weighting favorable and unfavorable interactions between the two molecules. Ligand-based scoring, on the other hand, focuses on how well the shape and chemistry of each ligand candidate overlay on a three-dimensional reference ligand. Our hypothesis is that a hybrid approach, using ligand-based scoring to rank dockings selected by protein-ligand scoring, can ensure that high-ranking molecules mimic the shape and chemistry of a known ligand while also complementing the binding site. Results from applying this approach to screen nearly 70 000 National Cancer Institute (NCI) compounds for thrombin inhibitors tend to support the hypothesis. EON ligand-based ranking of docked molecules yielded the majority (4/5) of newly discovered, low to mid-micromolar inhibitors from a panel of 27 assayed compounds, whereas ranking docked compounds by protein-ligand scoring alone resulted in one new inhibitor. Since the results depend on the choice of scoring function, an analysis of properties was performed on the top-scoring docked compounds according to five different protein-ligand scoring functions, plus EON scoring using three different reference compounds. The results indicate that the choice of scoring function, even among scoring functions measuring the same types of interactions, can have an unexpectedly large effect on which compounds are chosen from screening. Furthermore, there was almost no overlap between the top-scoring compounds from protein-ligand versus ligand-based scoring, indicating the two approaches provide complementary information. Matchprint analysis, a new addition to the SLIDE (Screening Ligands by Induced-fit Docking, Efficiently) screening toolset, facilitated comparison of docked molecules' interactions with those of known inhibitors. The majority of interactions conserved among top-scoring compounds for a given scoring function, and from the different scoring functions, proved to be conserved interactions in known inhibitors. This was particularly true in the S1 pocket, which was occupied by all the docked compounds.

• PLATINUM: a web tool for analysis of hydrophobic/hydrophilic organization of biomolecular complexes.
Pyrkov, Timothy V and Chugunov, Anton O and Krylov, Nikolay A and Nolde, Dmitry E and Efremov, Roman G
Bioinformatics (Oxford, England), 2009, 25(9), 1201-1202
PMID: 19244385     doi: 10.1093/bioinformatics/btp111

The PLATINUM (Protein-Ligand ATtractions Investigation NUMerically) web service is designed for analysis and visualization of hydrophobic/hydrophilic properties of biomolecules supplied as 3D-structures. Furthermore, PLATINUM provides a number of tools for quantitative characterization of the hydrophobic/hydrophilic match in biomolecular complexes e.g. in docking poses. These complement standard scoring functions. The calculations are based on the concept of empirical Molecular Hydrophobicity Potential (MHP). AVAILABILITY: The PLATINUM web tool as well as detailed documentation and tutorial are available free of charge for academic users at http://model.nmr.ru/platinum/. PLATINUM requires Java 5 or higher and Adobe Flash Player 9. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

• Docking ligands into flexible and solvated macromolecules. 4. Are popular scoring functions accurate for this class of proteins?
Englebienne, Pablo and Moitessier, Nicolas
Journal of chemical information and modeling, 2009, 49(6), 1568-1580
PMID: 19445499     doi: 10.1021/ci8004308

In our previous report, we investigated the impact of protein flexibility and the presence of water molecules on the pose-prediction accuracy of major docking programs. To complete these investigations, we report herein a study of the impact of these two aspects on the accuracy of scoring functions. To this effect, we developed two sets of protein/ligand complexes made up of ligands cross-docked or cocrystallized with a large variety of proteins, featuring bridging water molecules and demonstrating protein flexibility. Efforts were made to reduce the correlation between the molecular weights of the selected ligands and their binding affinities, a major bias in some previously reported benchmark sets. Using these sets, 18 available scoring functions have been assessed for their accuracy to predict binding affinities and to rank-order compounds by their affinity to cocrystallized proteins. This study confirmed the good and similar accuracy of Xscore, GlideScore, DrugScore(CSD), GoldScore, PLP1, ChemScore, RankScore, and the eHiTS scoring function. Our next investigations demonstrated that most of the assessed scoring functions were much less accurate when the correct protein conformation was not provided. This study also revealed that considering the water molecules for scoring does not greatly affect the accuracy. Finally, this work sheds light on the high correlation between scoring functions and the poor increase in accuracy one can expect from consensus scoring.

• Docking ligands into flexible and solvated macromolecules. 5. Force-field-based prediction of binding affinities of ligands to proteins.
Englebienne, Pablo and Moitessier, Nicolas
Journal of chemical information and modeling, 2009, 49(11), 2564-2571
PMID: 19928836     doi: 10.1021/ci900251k

We report herein our efforts in the development of three empirical scoring functions with application in protein-ligand docking. A first scoring function was developed from 209 crystal structures of protein-ligand complexes and a second one from 946 cross-docked complexes. Tuning of the coefficients for the different terms making up these functions was performed by an iterative approach to optimize the correlations between observed activities and calculated scores. A third scoring function was developed from libraries of known actives and decoys docked to six different protein conformational ensembles. In the latter case, the tuning of the coefficients was performed so as to optimize the area under the curve of a receiver operating characteristic (ROC) for the discrimination of actives and inactives. The newly developed scoring functions were next assessed on independent sets of protein-ligand complexes for their ability to predict binding affinities and to discriminate actives from inactives. In the first validation the first function, which was trained on active compounds only, performed as well as other commonly used ones. On a high-throughput virtual screening validation on five protein conformational ensembles, the third scoring function that included data from inactive compounds performed significantly better. This validation showed that the inclusion of data from inactive compounds is critical for performance in virtual high-throughput screening applications.

• Docking to heme proteins.
Röhrig, Ute F and Grosdidier, Aurélien and Zoete, Vincent and Michielin, Olivier
Journal of computational chemistry, 2009, 30(14), 2305-2315
PMID: 19288474     doi: 10.1002/jcc.21244

In silico screening has become a valuable tool in drug design, but some drug targets represent real challenges for docking algorithms. This is especially true for metalloproteins, whose interactions with ligands are difficult to parametrize. Our docking algorithm, EADock, is based on the CHARMM force field, which assures a physically sound scoring function and a good transferability to a wide range of systems, but also exhibits difficulties in case of some metalloproteins. Here, we consider the therapeutically important case of heme proteins featuring an iron core at the active site. Using a standard docking protocol, where the iron-ligand interaction is underestimated, we obtained a success rate of 28% for a test set of 50 heme-containing complexes with iron-ligand contact. By introducing Morse-like metal binding potentials (MMBP), which are fitted to reproduce density functional theory calculations, we are able to increase the success rate to 62%. The remaining failures are mainly due to specific ligand-water interactions in the X-ray structures. Testing of the MMBP on a second data set of non iron binders (14 cases) demonstrates that they do not introduce a spurious bias towards metal binding, which suggests that they may reliably be used also for cross-docking studies.

• SeleX-CS: a new consensus scoring algorithm for hit discovery and lead optimization.
Bar-Haim, Shay and Aharon, Ayelet and Ben-Moshe, Tal and Marantz, Yael and Senderowitz, Hanoch
Journal of chemical information and modeling, 2009, 49(3), 623-633
PMID: 19231809     doi: 10.1021/ci800335j

Identifying active compounds (hits) that bind to biological targets of pharmaceutical relevance is the cornerstone of drug design efforts. Structure based virtual screening, namely, the in silico evaluation of binding energies and geometries between a protein and its putative ligands, has emerged over the past few years as a promising approach in this field. The success of the method relies on the availability of reliable 3-dimensional (3D) structures of the target protein and its candidate ligands (the screening library), a reliable docking method that can fit the different ligands into the protein's binding site, and an accurate scoring function that can rank the resulting binding modes in accord with their binding affinities. This last requirement is arguably the most difficult to meet due to the complexity of the binding process. A potential solution to this so-called scoring problem is the usage of multiple scoring functions in an approach known as consensus scoring. Several consensus scoring methods were suggested in the literature and have generally demonstrated an improved ranking of screening libraries relative to individual scoring functions. Nevertheless, current consensus scoring strategies suffer from several shortcomings, in particular, strong dependence on the initial parameters and an incomplete treatment of inactive compounds. In this work we present a new consensus scoring algorithm (SeleX-Consensus Scoring abbreviated to SeleX-CS) specifically designed to address these limitations: (i) A subset of the initial set of the scoring functions is allowed to form the consensus score, and this subset is optimized via a Monte Carlo/Simulated Annealing procedure. (ii) Rank redundancy between the members of the screening library is removed. (iii) The method explicitly considers the presence of inactive compounds. The new algorithm was applied to the ranking of screening libraries targeting two G-protein coupled receptors (GPCR). Excellent enrichment factors were obtained in both cases: For the cannabinoid receptor 1 (CB1), SeleX-CS outperformed the best single score and afforded an enrichment factor of 41 at 1% of the screening library compared with the best single score value of 15 (GOLD_Fitness). For the chemokine receptor type 2 (CCR2) SeleX-CS afforded an enrichment factor of 72 (again at 1% of the screening library) once more outperforming any single score (enrichment factor of 20 by GSCORE). Moreover, SeleX-CS demonstrated success rates of 67% (CCR2) and 73% (CB1) when applied to ranking an external test set. In both cases, the new algorithm also afforded good derichment of inactive compounds (i.e., the ability to push inactive compounds to the bottom of the ranked library). The method was then extended to rank a lead optimization series targeting the Kv4.3 potassium ion channel, resulting in a Spearman's correlation coefficient, p

• Comparative assessment of scoring functions on a diverse test set.
Cheng, Tiejun and Li, Xun and Li, Yan and Liu, Zhihai and Wang, Renxiao
Journal of chemical information and modeling, 2009, 49(4), 1079-1093
PMID: 19358517     doi: 10.1021/ci9000053

Scoring functions are widely applied to the evaluation of protein-ligand binding in structure-based drug design. We have conducted a comparative assessment of 16 popular scoring functions implemented in main-stream commercial software or released by academic research groups. A set of 195 diverse protein-ligand complexes with high-resolution crystal structures and reliable binding constants were selected through a systematic nonredundant sampling of the PDBbind database and used as the primary test set in our study. All scoring functions were evaluated in three aspects, that is, "docking power", "ranking power", and "scoring power", and all evaluations were independent from the context of molecular docking or virtual screening. As for "docking power", six scoring functions, including GOLD::ASP, DS::PLP1, DrugScore(PDB), GlideScore-SP, DS::LigScore, and GOLD::ChemScore, achieved success rates over 70% when the acceptance cutoff was root-mean-square deviation < 2.0 A. Combining these scoring functions into consensus scoring schemes improved the success rates to 80% or even higher. As for "ranking power" and "scoring power", the top four scoring functions on the primary test set were X-Score, DrugScore(CSD), DS::PLP, and SYBYL::ChemScore. They were able to correctly rank the protein-ligand complexes containing the same type of protein with success rates around 50%. Correlation coefficients between the experimental binding constants and the binding scores computed by these scoring functions ranged from 0.545 to 0.644. Besides the primary test set, each scoring function was also tested on four additional test sets, each consisting of a certain number of protein-ligand complexes containing one particular type of protein. Our study serves as an updated benchmark for evaluating the general performance of today's scoring functions. Our results indicate that no single scoring function consistently outperforms others in all three aspects. Thus, it is important in practice to choose the appropriate scoring functions for different purposes.

• Testing assumptions and hypotheses for rescoring success in protein-ligand docking.
O'Boyle, Noel M. and Liebeschuetz, John W and Cole, Jason C
Journal of chemical information and modeling, 2009, 49(8), 1871-1878
PMID: 19645429     doi: 10.1021/ci900164f

In protein-ligand docking, the scoring function is responsible for identifying the correct pose of a particular ligand as well as separating ligands from nonligands. Recently there has been considerable interest in schemes that combine results from several scoring functions in an effort to achieve improved performance in virtual screens. One such scheme is consensus scoring, which involves combining the results from several rescoring experiments. Although there have been a number of studies that have investigated factors affecting success in consensus scoring, these studies have not addressed the question of why a rescoring strategy works in the first place. Here we propose and test two alternative hypotheses for why rescoring has the potential to improve results, using GOLD 4.0. The "consensus" hypothesis is that rescoring is a way of combining results from two scoring functions such that only true positives are likely to score highly. The "complementary" hypothesis is that the two scoring functions used in rescoring have complementary strengths; one is better at ranking actives with respect to inactives while the other is better at ranking poses of actives. We find that in general it is this hypothesis that explains success in a rescoring experiment. We also test an assumption of any rescoring method, which is that the scores obtained are representative of the fitness of the docked pose. We find that although rescored poses tended to have slightly higher clash values than their docked equivalents, in general the scores were representative.

• Empirical scoring functions for advanced protein-ligand docking with PLANTS.
Korb, Oliver and Stutzle, Thomas and Exner, Thomas E.
Journal of chemical information and modeling, 2009, 49(1), 84-96
PMID: 19125657     doi: 10.1021/ci800298z

In this paper we present two empirical scoring functions, PLANTS(CHEMPLP) and PLANTS(PLP), designed for our docking algorithm PLANTS (Protein-Ligand ANT System), which is based on ant colony optimization (ACO). They are related, regarding their functional form, to parts of already published scoring functions and force fields. The parametrization procedure described here was able to identify several parameter settings showing an excellent performance for the task of pose prediction on two test sets comprising 298 complexes in total. Up to 87% of the complexes of the Astex diverse set and 77% of the CCDC/Astex clean listnc (noncovalently bound complexes of the clean list) could be reproduced with root-mean-square deviations of less than 2 A with respect to the experimentally determined structures. A comparison with the state-of-the-art docking tool GOLD clearly shows that this is, especially for the druglike Astex diverse set, an improvement in pose prediction performance. Additionally, optimized parameter settings for the search algorithm were identified, which can be used to balance pose prediction reliability and search speed.

## 2008

• Bootstrap-based consensus scoring method for protein-ligand docking.
Journal of chemical information and modeling, 2008, 48(5), 988-996
PMID: 18426197     doi: 10.1021/ci700204v

To improve the performance of a single scoring function used in a protein-ligand docking program, we developed a bootstrap-based consensus scoring (BBCS) method, which is based on ensemble learning. BBCS combines multiple scorings, each of which has the same function form but different energy-parameter sets. These multiple energy-parameter sets are generated in two steps: (1) generation of training sets by a bootstrap method and (2) optimization of energy-parameter set by a Z-score approach, which is based on energy landscape theory as used in protein folding, against each training set. In this study, we applied BBCS to the FlexX scoring function. Using given 50 complexes, we generated 100 training sets and obtained 100 optimized energy-parameter sets. These parameter sets were tested against 48 complexes different from the training sets. BBCS was shown to be an improvement over single scoring when using a parameter set optimized by the same Z-score approach. Comparing BBCS with the original FlexX scoring function, we found that (1) the success rate of recognizing the crystal structure at the top relative to decoys increased from 33.3% to 52.1% and that (2) the rank of the crystal structure improved for 54.2% of the complexes and worsened for none. We also found that BBCS performed better than conventional consensus scoring (CS).

• Q-Dock: Low-resolution flexible ligand docking with pocket-specific threading restraints.
Brylinski, Michal and Skolnick, Jeffrey
Journal of computational chemistry, 2008, 29(10), 1574-1588
PMID: 18293308     doi: 10.1002/jcc.20917

The rapidly growing number of theoretically predicted protein structures requires robust methods that can utilize low-quality receptor structures as targets for ligand docking. Typically, docking accuracy falls off dramatically when apo or modeled receptors are used in docking experiments. Low-resolution ligand docking techniques have been developed to deal with structural inaccuracies in predicted receptor models. In this spirit, we describe the development and optimization of a knowledge-based potential implemented in Q-Dock, a low-resolution flexible ligand docking approach. Self-docking experiments using crystal structures reveals satisfactory accuracy, comparable with all-atom docking. All-atom models reconstructed from Q-Dock's low-resolution models can be further refined by even a simple all-atom energy minimization. In decoy-docking against distorted receptor models with a root-mean-square deviation, RMSD, from native of approximately 3 A, Q-Dock recovers on average 15-20% more specific contacts and 25-35% more binding residues than all-atom methods. To further improve docking accuracy against low-quality protein models, we propose a pocket-specific protein-ligand interaction potential derived from weakly homologous threading holo-templates. The success rate of Q-Dock employing a pocket-specific potential is 6.3 times higher than that previously reported for the Dolores method, another low-resolution docking approach.

• Consensus scoring with feature selection for structure-based virtual screening
Teramoto, Reiji and Fukunishi, Hiroaki
Journal of chemical information and modeling, 2008, 48(2), 288-295
doi: 10.1021/ci700239t

The evaluation of ligand conformations is a crucial aspect of structure-based virtual screening, and scoring functions play significant roles in it. While consensus scoring (CS) generally improves enrichment by compensating for the deficiencies of each scoring function, the strategy of how individual scoring functions are selected remains a challenging task when few known active compounds are available. To address this problem, we propose feature selection-based consensus scoring (FSCS), which performs supervised feature selection with docked native ligand conformations to select complementary scoring functions. We evaluated the enrichments of five scoring functions (F-Score, D-Score, PMF, G-Score, and ChemScore), FSCS, and RCS (rank-by-rank consensus scoring) for four different target proteins: acetylcholine esterase (AChE), thrombin (thrombin), phosphodiesterase 5 (PDE5), and peroxisome proliferator-activated receptor gamma (PPAR gamma). The results indicated that FSCS was able to select the complementary scoring functions and enhance ligand enrichments and that it outperformed RCS and the individual scoring functions for all target proteins. They also indicated that the performances of the single scoring functions were strongly dependent on the target protein. An especially favorable result with implications for practical drug screening is that FSCS performs well even if only one 3D structure of the protein-ligand complex is known. Moreover, we found that one can infer which scoring functions significantly enrich active compounds by using feature selection before actual docking and that the selected scoring functions are complementary.

• Using buriedness to improve discrimination between actives and inactives in docking
O'Boyle, Noel M. and Brewerton, Suzanne C. and Taylor, Robin
Journal of chemical information and modeling, 2008, 48(6), 1269-1278
PMID: 18533645     doi: 10.1021/ci8000452

A continuing problem in protein-ligand docking is the correct relative ranking of active molecules versus inactives. Using the ChemScore scoring function as implemented in the GOLD docking software, we have investigated the effect of scaling hydrogen bond, metal-ligand, and lipophilic interactions based on the buriedness of the interaction. Buriedness was measured using the receptor density, the number of protein heavy atoms within 8.0 angstrom. Terms in the scaling functions were optimized using negative data, represented by docked poses of inactive molecules. The objective function was the mean rank of the scores of the active poses in the Astex Diverse Set (Hartshorn et al. J. Med. Chem., 2007, 50, 726) with respect to the docked poses of 99 inactives. The final four-parameter model gave a substantial improvement in the average rank from 18.6 to 12.5. Similar results were obtained for an independent test set. Receptor density scaling is available as an option in the recent GOLD release.

• MedusaScore: an accurate force field-based scoring function for virtual drug screening.
Yin, Shuangye and Biedermannova, Lada and Vondrasek, Jiri and Dokholyan, Nikolay V
Journal of chemical information and modeling, 2008, 48(8), 1656-1662
PMID: 18672869     doi: 10.1021/ci8001167

Virtual screening is becoming an important tool for drug discovery. However, the application of virtual screening has been limited by the lack of accurate scoring functions. Here, we present a novel scoring function, MedusaScore, for evaluating protein-ligand binding. MedusaScore is based on models of physical interactions that include van der Waals, solvation, and hydrogen bonding energies. To ensure the best transferability of the scoring function, we do not use any protein-ligand experimental data for parameter training. We then test the MedusaScore for docking decoy recognition and binding affinity prediction and find superior performance compared to other widely used scoring functions. Statistical analysis indicates that one source of inaccuracy of MedusaScore may arise from the unaccounted entropic loss upon ligand binding, which suggests avenues of approach for further MedusaScore improvement.

## 2007

• Supervised consensus scoring for docking and virtual screening
Teramoto, Reiji and Fukunishi, Hiroaki
Journal of chemical information and modeling, 2007, 47(2), 526-534
doi: 10.1021/ci6004993

Docking programs are widely used to discover novel ligands efficiently and can predict protein-ligand complex structures with reasonable accuracy and speed. However, there is an emerging demand for better performance from the scoring methods. Consensus scoring (CS) methods improve the performance by compensating for the deficiencies of each scoring function. However, conventional CS and existing scoring functions have the same problems, such as a lack of protein flexibility, inadequate treatment of salvation, and the simplistic nature of the energy function used. Although there are many problems in current scoring functions, we focus our attention on the incorporation of unbound ligand conformations. To address this problem, we propose supervised consensus scoring (SCS), which takes into account protein-ligand binding process using unbound ligand conformations with supervised learning. An evaluation of docking accuracy for 100 diverse protein-ligand complexes shows that SCS outperforms both CS and 11 scoring functions (PLP, F-Score, LigScore, DrugScore, LUDI, X-Score, AutoDock, PMF, G-Score, ChemScore, and D-score). The success rates of SCS range from 89% to 91% in the range of rmsd < 2 A, while those of CS range from 80% to 85%, and those of the scoring functions range from 26% to 76%. Moreover, we also introduce a method for judging whether a compound is active or inactive with the appropriate criterion for virtual screening. SCS performs quite well in docking accuracy and is presumably useful for screening large-scale compound databases before predicting binding affinity.

• Solvated interaction energy (SIE) for scoring protein-ligand binding affinities. 1. Exploring the parameter space.
Naïm, Marwen and Bhat, Sathesh and Rankin, Kathryn N and Dennis, Sheldon and Chowdhury, Shafinaz F and Siddiqi, Imran and Drabik, Piotr and Sulea, Traian and Bayly, Christopher I and Jakalian, Araz and Purisima, Enrico O
Journal of chemical information and modeling, 2007, 47(1), 122-133
PMID: 17238257     doi: 10.1021/ci600406v

We present a binding free energy function that consists of force field terms supplemented by solvation terms. We used this function to calibrate the solvation model along with the binding interaction terms in a self-consistent manner. The motivation for this approach was that the solute dielectric-constant dependence of calculated hydration gas-to-water transfer free energies is markedly different from that of binding free energies (J. Comput. Chem. 2003, 24, 954). Hence, we sought to calibrate directly the solvation terms in the context of a binding calculation. The five parameters of the model were systematically scanned to best reproduce the absolute binding free energies for a set of 99 protein-ligand complexes. We obtained a mean unsigned error of 1.29 kcal/mol for the predicted absolute binding affinity in a parameter space that was fairly shallow near the optimum. The lowest errors were obtained with solute dielectric values of Din

• Supervised scoring models with docked ligand conformations for structure-based virtual screening.
Teramoto, Reiji and Fukunishi, Hiroaki
Journal of chemical information and modeling, 2007, 47(5), 1858-1867
PMID: 17685604     doi: 10.1021/ci700116z

Protein-ligand docking programs have been used to efficiently discover novel ligands for target proteins from large-scale compound databases. However, better scoring methods are needed. Generally, scoring functions are optimized by means of various techniques that affect their fitness for reproducing X-ray structures and protein-ligand binding affinities. However, these scoring functions do not always work well for all target proteins. A scoring function should be optimized for a target protein to enhance enrichment for structure-based virtual screening. To address this problem, we propose the supervised scoring model (SSM), which takes into account the protein-ligand binding process using docked ligand conformations with supervised learning for optimizing scoring functions against a target protein. SSM employs a rough linear correlation between binding free energy and the root mean square deviation of a native ligand for predicting binding energy. We applied SSM to the FlexX scoring function, that is, F-Score, with five different target proteins: thymidine kinase (TK), estrogen receptor (ER), acetylcholine esterase (AChE), phosphodiesterase 5 (PDE5), and peroxisome proliferator-activated receptor gamma (PPARgamma). For these five proteins, SSM always enhanced enrichment better than F-Score, exhibiting superior performance that was particularly remarkable for TK, AChE, and PPARgamma. We also demonstrated that SSM is especially good at enhancing enrichments of the top ranks of screened compounds, which is useful in practical drug screening.

## 2006

• M-Score: A Knowledge-Based Potential Scoring Function Accounting for Protein Atom Mobility
Yang, Chao-Yie and Wang, Renxiao and Wang, Shaomeng
Journal of medicinal chemistry, 2006, 49(20), 5903-5911
doi: 10.1021/jm050043w

• Information-driven protein-DNA docking using HADDOCK: it is a matter of flexibility.
van Dijk, Marc and van Dijk, Aalt D J and Hsu, Victor and Boelens, Rolf and Bonvin, Alexandre M J J
Nucleic acids research, 2006, 34(11), 3317-3325
PMID: 16820531     doi: 10.1093/nar/gkl412

Intrinsic flexibility of DNA has hampered the development of efficient protein-DNA docking methods. In this study we extend HADDOCK (High Ambiguity Driven DOCKing) [C. Dominguez, R. Boelens and A. M. J. J. Bonvin (2003) J. Am. Chem. Soc. 125, 1731-1737] to explicitly deal with DNA flexibility. HADDOCK uses non-structural experimental data to drive the docking during a rigid-body energy minimization, and semi-flexible and water refinement stages. The latter allow for flexibility of all DNA nucleotides and the residues of the protein at the predicted interface. We evaluated our approach on the monomeric repressor-DNA complexes formed by bacteriophage 434 Cro, the Escherichia coli Lac headpiece and bacteriophage P22 Arc. Starting from unbound proteins and canonical B-DNA we correctly predict the correct spatial disposition of the complexes and the specific conformation of the DNA in the published complexes. This information is subsequently used to generate a library of pre-bent and twisted DNA structures that served as input for a second docking round. The resulting top ranking solutions exhibit high similarity to the published complexes in terms of root mean square deviations, intermolecular contacts and DNA conformation. Our two-stage docking method is thus able to successfully predict protein-DNA complexes from unbound constituents using non-structural experimental data to drive the docking.

• A method for induced-fit docking, scoring, and ranking of flexible ligands. Application to peptidic and pseudopeptidic beta-secretase (BACE 1) inhibitors.
Moitessier, Nicolas and Therrien, Eric and Hanessian, Stephen
Journal of medicinal chemistry, 2006, 49(20), 5885-5894
PMID: 17004704     doi: 10.1021/jm050138y

Inhibition of beta-secretase (BACE 1) has recently been investigated as a promising therapeutic approach in the treatment of Alzheimer's disease, and a growing number of BACE 1 inhibitors and crystal structures of BACE 1/inhibitors complexes have been reported. We report herein a predictive computational method and its application to potential BACE 1 inhibitors. Using a training set of 50 known highly flexible inhibitors, we developed a docking method that accounts for the flexibility of both the protein and the inhibitors. Protein flexibility is accounted for using a specifically designed genetic algorithm. We next developed a scoring function consisting of force field evaluation of the inhibitor/protein interactions and two additional terms for hydrogen bonding and entropy change upon binding. Discarding three outliers from the training set, our protocol was found to perform well with an rmsd of 1.19 kcal/mol. Evaluation of the predictive power was next carried out by virtual screening of 80 synthetic compounds. The significant enrichment at the top of the ranking list in active compounds demonstrated the ability of the docking and scoring protocol to rank the compounds relative to their activities.

• Scoring functions for protein-ligand docking.
Jain, Ajay N
Current Protein & Peptide Science, 2006, 7(5), 407-420
PMID: 17073693

Virtual screening by molecular docking has become established as a method for drug lead discovery and optimization. All docking algorithms make use of a scoring function in combination with a method of search. Two theoretical aspects of scoring function performance dominate operational performance. The first is the degree to which a scoring function has a global extremum within the ligand pose landscape at the proper location. The second is the degree to which the magnitude of the function at the extremum is accurate. Presuming adequate search strategies, a scoring function's location performance will dominate behavior with respect to docking accuracy: the degree to which a predicted pose of a ligand matches experimental observation. A scoring function's magnitude performance will dominate behavior with respect to screening utility: enrichment of true ligands over non-ligands. Magnitude estimation also controls pure scoring accuracy: the degree to which bona fide ligands of a particular protein may be correctly ranked. Approaches to the development of scoring functions have varied widely, with a number of functions yielding similarly high levels of performance relating to the location issue. However, even among functions performing equally well on location, widely varying performance is observed on the question of magnitude. In many cases, performance is good enough to yield high enrichments of true ligands versus non-ligands in screening across a wide variety of protein types. Generally, performance is not good enough to correctly rank among true ligands. Strategies for improvement are discussed.

• Parameter estimation for scoring protein-ligand interactions using negative training data.
Pham, Tuan A and Jain, Ajay N
Journal of medicinal chemistry, 2006, 49(20), 5856-5868
PMID: 17004701     doi: 10.1021/jm050040j

Surflex-Dock employs an empirically derived scoring function to rank putative protein-ligand interactions by flexible docking of small molecules to proteins of known structure. The scoring function employed by Surflex was developed purely on the basis of positive data, comprising noncovalent protein-ligand complexes with known binding affinities. Consequently, scoring function terms for improper interactions received little weight in parameter estimation, and an ad hoc scheme for avoiding protein-ligand interpenetration was adopted. We present a generalized method for incorporating synthetically generated negative training data, which allows for rigorous estimation of all scoring function parameters. Geometric docking accuracy remained excellent under the new parametrization. In addition, a test of screening utility covering a diverse set of 29 proteins and corresponding ligand sets showed improved performance. Maximal enrichment of true ligands over nonligands exceeded 20-fold in over 80% of cases, with enrichment of greater than 100-fold in over 50% of cases.

• PSI-DOCK: towards highly efficient and accurate flexible ligand docking.
Pei, Jianfeng and Wang, Qi and Liu, Zhenming and Li, Qingliang and Yang, Kun and Lai, Luhua
Proteins, 2006, 62(4), 934-946
PMID: 16395666     doi: 10.1002/prot.20790

We have developed a new docking method, Pose-Sensitive Inclined (PSI)-DOCK, for flexible ligand docking. An improved SCORE function has been developed and used in PSI-DOCK for binding free energy evaluation. The improved SCORE function was able to reproduce the absolute binding free energies of a training set of 200 protein-ligand complexes with a correlation coefficient of 0.788 and a standard error of 8.13 kJ/mol. For ligand binding pose exploration, a unique searching strategy was designed in PSI-DOCK. In the first step, a tabu-enhanced genetic algorithm with a rapid shape-complementary scoring function is used to roughly explore and store potential binding poses of the ligand. Then, these predicted binding poses are optimized and compete against each other by using a genetic algorithm with the accurate SCORE function to determine the binding pose with the lowest docking energy. The PSI-DOCK 1.0 program is highly efficient in identifying the experimental binding pose. For a test dataset of 194 complexes, PSI-DOCK 1.0 achieved a 67% success rate (RMSD < 2.0 A) for only one run and a 74% success rate for 10 runs. PSI-DOCK can also predict the docking binding free energy with high accuracy. For a test set of 64 complexes, the correlation between the experimentally observed binding free energies and the docking binding free energies for 64 complexes is r

• eHITS: An innovative approach to the docking and scoring function problems
Zsoldos, Zsolt and Reid, Darryl and Simon, Aniko and Sadjad, Bashir S and Johnson, A. Peter
Current Protein & Peptide Science, 2006, 7(5), 421-435
PMID: 17073694

Virtual Ligand Screening (VLS) has become an integral part of the drug design process for many pharmaceutical companies. In protein structure based VLS the aim is to find a ligand that has a high binding affinity to the target receptor whose 3D structure is known. This review will describe the docking tool eHiTS. eHiTS is an exhaustive and systematic docking tool which contains many automated features that simplify the drug design workflow. A description of the unique docking algorithm and novel approach to scoring used within eHiTS is presented. In addition a validation study is presented that demonstrates the accuracy and wide applicability of eHiTs in re-docking bound ligands into their receptors.

• Prediction of Protein−Ligand Interactions. Docking and Scoring: Successes and Gaps
Leach, Andrew R and Shoichet, Brian K and Peishoff, Catherine E
Journal of medicinal chemistry, 2006, 49(20), 5851-5855
doi: 10.1021/jm060999m

Computational methods have become standard in today's medicinal chemistry tool kit. Like any tool, it is important to periodically evaluate utility and ask how function can be improved. In this section of the Journal, we call attention to the area of calculating molecular interactions, specifically docking, the positioning of a ligand in a protein binding site, and scoring, the quality assessment of docked ligands. As several recent reviews have made clear,1-3 the technology has been productive for both finding and elaborating bioactive molecules. But has docking and scoring delivered on the promises first made over 20 years ago? To consider that question, we follow up on an extensive symposium held in Philadelphia during the 2004 Fall National Meeting of the American Chemistry Society and on subsequent meetings sponsored by the National Institutes of Health (NIH) and the National Institute of Standards and Technology (NIST) in 2005 and 2006 to address the outcomes of the American Chemical Society symposium. Speakers at the symposium were invited to contribute original manuscripts to be published with this overview to highlight the area of docking and scoring and to identify some of the major gaps yet to be addressed.

• Prediction of Protein−Ligand Interactions. Docking and Scoring: Successes and Gaps
Leach, Andrew R and Shoichet, Brian K and Peishoff, Catherine E
Journal of medicinal chemistry, 2006, 49(20), 5851-5855
doi: 10.1021/jm060999m

Computational methods have become standard in today's medicinal chemistry tool kit. Like any tool, it is important to periodically evaluate utility and ask how function can be improved. In this section of the Journal, we call attention to the area of calculating molecular interactions, specifically docking, the positioning of a ligand in a protein binding site, and scoring, the quality assessment of docked ligands. As several recent reviews have made clear,1-3 the technology has been productive for both finding and elaborating bioactive molecules. But has docking and scoring delivered on the promises first made over 20 years ago? To consider that question, we follow up on an extensive symposium held in Philadelphia during the 2004 Fall National Meeting of the American Chemistry Society and on subsequent meetings sponsored by the National Institutes of Health (NIH) and the National Institute of Standards and Technology (NIST) in 2005 and 2006 to address the outcomes of the American Chemical Society symposium. Speakers at the symposium were invited to contribute original manuscripts to be published with this overview to highlight the area of docking and scoring and to identify some of the major gaps yet to be addressed.

## 2005

• Validation and use of the MM-PBSA approach for drug discovery.
Kuhn, Bernd and Gerber, Paul and Schulz-Gasch, Tanja and Stahl, Martin
Journal of medicinal chemistry, 2005, 48(12), 4040-4048
PMID: 15943477     doi: 10.1021/jm049081q

The MM-PBSA approach has become a popular method for calculating binding affinities of biomolecular complexes. Published application examples focus on small test sets and few proteins and, hence, are of limited relevance in assessing the general validity of this method. To further characterize MM-PBSA, we report on a more extensive study involving a large number of ligands and eight different proteins. Our results show that applying the MM-PBSA energy function to a single, relaxed complex structure is an adequate and sometimes more accurate approach than the standard free energy averaging over molecular dynamics snapshots. The use of MM-PBSA on a single structure is shown to be valuable (a) as a postdocking filter in further enriching virtual screening results, (b) as a helpful tool to prioritize de novo design solutions, and (c) for distinguishing between good and weak binders (DeltapIC(50) > or

• DrugScore(CSD)-knowledge-based scoring function derived from small molecule crystal data with superior recognition rate of near-native ligand poses and better affinity prediction.
Velec, Hans F G and Gohlke, Holger and Klebe, Gerhard
Journal of medicinal chemistry, 2005, 48(20), 6296-6303
PMID: 16190756     doi: 10.1021/jm050436v

Following the formalism used for the development of the knowledge-based scoring function DrugScore, new distance-dependent pair potentials are obtained from nonbonded interactions in small organic molecule crystal packings. Compared to potentials derived from protein-ligand complexes, the better resolved small molecule structures provide relevant contact data in a more balanced distribution of atom types and produce potentials of superior statistical significance and more detailed shape. Applied to recognizing binding geometries of ligands docked into proteins, this new scoring function (DrugScore(CSD)) ranks the crystal structures of 100 protein-ligand complexes best among up to 100 generated decoy geometries in 77% of all cases. Accepting root-mean-square deviations (rmsd) of up to 2 angstroms from the native pose as well-docked solutions, a correct binding mode is found in 87% of the cases. This translates into an improvement of the new scoring function of 57% with respect to the retrieval of the crystal structure and 20% with respect to the identification of a well-docked ligand pose compared to the original Protein Data Bank-based DrugScore. In the analysis of decoy geometries of cross-docking studies, DrugScore(CSD) shows equivalent or increased performance compared to the original PDB-based DrugScore. Furthermore, DrugScore(CSD) predicts binding affinities convincingly. Reducing the set of docking solutions to examples that deviate increasingly from the native pose results in a loss of performance of DrugScore(CSD). This indicates that a necessary prerequisite to successfully resolving the scoring problem with a more discriminative scoring function is the generation of highly accurate ligand poses, which approximate the native pose to below 1 angstroms rmsd, in a docking run.

• LigScore: a novel scoring function for predicting binding affinities.
Krammer, André and Kirchhoff, Paul D and Jiang, X and Venkatachalam, C M and Waldman, Marvin
Journal of molecular graphics & modelling, 2005, 23(5), 395-407
PMID: 15781182     doi: 10.1016/j.jmgm.2004.11.007

We present two new empirical scoring functions, LigScore1 and LigScore2, that attempt to accurately predict the binding affinity between ligand molecules and their protein receptors. The LigScore functions consist of three distinct terms that describe the van der Waals interaction, the polar attraction between the ligand and protein, and the desolvation penalty attributed to the binding of the polar ligand atoms to the protein and vice versa. Utilizing a regression approach on a data set of 118 protein-ligand complexes we have obtained a linear equation, LigScore2, using these three descriptors. LigScore2 has good predictability with regard to experimental pKi values yielding a correlation coefficient, r2), of 0.75 and a standard deviation of 1.04 over the training data set, which consists of a diverse set of proteins that span more than seven protein families.

## 2004

• Docking and scoring in virtual screening for drug discovery: methods and applications.
Kitchen, Douglas B and Decornez, Hélène and Furr, John R and Bajorath, Jürgen
Nature reviews. Drug discovery, 2004, 3(11), 935-949
PMID: 15520816     doi: 10.1038/nrd1549

Computational approaches that 'dock' small molecules into the structures of macromolecular targets and 'score' their potential complementarity to binding sites are widely used in hit identification and lead optimization. Indeed, there are now a number of drugs whose development was heavily influenced by or based on structure-based design and screening strategies, such as HIV protease inhibitors. Nevertheless, there remain significant challenges in the application of these approaches, in particular in relation to current scoring schemes. Here, we review key concepts and specific features of small-molecule-protein docking methods, highlight selected applications and discuss recent advances that aim to address the acknowledged limitations of established approaches.

• Scoring functions for protein-ligand interactions: a critical perspective
Schulz-Gasch, T
Drug Discovery Today: Technologies, 2004, 1(3), 231-239

Scoring functions play an essential role in structure- based virtual screening. They are required to guide the docking of candidate compounds to structures of receptor binding sites, to select probable binding modes, and to discriminate binders from non-binders. Although many scoring functions have successfully been used to identify novel ligands for a wide variety of targets, much work remains to be done to avoid incorrect prediction of binding modes and high num- bers of false positives. This review gives an overview of the current state of the field and outlines key issues for the further development of scoring functions.

• Native atom types for knowledge-based potentials: application to binding energy prediction.
Dominy, Brian N and Shakhnovich, Eugene I
Journal of medicinal chemistry, 2004, 47(18), 4538-4558
PMID: 15317465     doi: 10.1021/jm0498046

Knowledge-based potentials have been found useful in a variety of biophysical studies of macromolecules. Recently, it has also been shown in self-consistent studies that it is possible to extract quantities consistent with pair potentials from model structural databases. In this study, we attempt to extend the results obtained from these self-consistent studies toward the extraction of realistic pair potentials from the Protein Data Bank (PDB). The new method utilizes a clustering approach to define atom types within the PDB consistent with the optimal effective pairwise potential. The method has been integrated into the SMoG drug design package, resulting in an improved approach for the rapid and accurate estimation of binding affinities from structural information. Using this approach, it is possible to generate simple knowledge-based potentials that correlate (R

• Validation of an empirical RNA-ligand scoring function for fast flexible docking using Ribodock.
Morley, S David and Afshar, Mohammad
Journal of computer-aided molecular design, 2004, 18(3), 189-208
PMID: 15368919

We report the design and validation of a fast empirical function for scoring RNA-ligand interactions, and describe its implementation within RiboDock, a virtual screening system for automated flexible docking. Building on well-known protein-ligand scoring function foundations, features were added to describe the interactions of common RNA-binding functional groups that were not handled adequately by conventional terms, to disfavour non-complementary polar contacts, and to control non-specific charged interactions. The results of validation experiments against known structures of RNA-ligand complexes compare favourably with previously reported methods. Binding modes were well predicted in most cases and good discrimination was achieved between native and non-native ligands for each binding site, and between native and non-native binding sites for each ligand. Further evidence of the ability of the method to identify true RNA binders is provided by compound selection ('enrichment factor') experiments based around a series of HIV-1 TAR RNA-binding ligands. Significant enrichment in true binders was achieved amongst high scoring docking hits, even when selection was from a library of structurally related, positively charged molecules. Coupled with a semi-automated cavity detection algorithm for identification of putative ligand binding sites, also described here, the method is suitable for the screening of very large databases of molecules against RNA and RNA-protein interfaces, such as those presented by the bacterial ribosome.

• Calculation of ligand-nucleic acid binding free energies with the generalized-born model in DOCK.
Kang, Xinshan and Shafer, Richard H and Kuntz, Irwin D
Biopolymers, 2004, 73(2), 192-204
PMID: 14755577     doi: 10.1002/bip.10541

The calculation of ligand-nucleic acid binding free energies is investigated by including solvation effects computed with the generalized-Born model. Modifications of the solvation module in DOCK, including introduction of all-atom parameters and revision of coefficients in front of different terms, are shown to improve calculations involving nucleic acids. This computing scheme is capable of calculating binding energies, with reasonable accuracy, for a wide variety of DNA-ligand complexes, RNA-ligand complexes, and even for the formation of double-stranded DNA. This implementation of GB/SA is also shown to be capable of discriminating strong ligands from poor ligands for a series of RNA aptamers without sacrificing the high efficiency of the previous implementation. These results validate this approach to screening large databases against nucleic acid targets.

• Evaluation and application of multiple scoring functions for a virtual screening experiment.
Xing, Li and Hodgkin, Edward and Liu, Qian and Sedlock, David
Journal of computer-aided molecular design, 2004, 18(5), 333-344
PMID: 15595460

In order to identify novel chemical classes of factor Xa inhibitors, five scoring functions (FlexX, DOCK, GOLD, ChemScore and PMF) were engaged to evaluate the multiple docking poses generated by FlexX. The compound collection was composed of confirmed potent factor Xa inhibitors and a subset of the LeadQuest screening compound library. Except for PMF the other four scoring functions succeeded in reproducing the crystal complex (PDB code: 1FAX). During virtual screening the highest hit rate (80%) was demonstrated by FlexX at an energy cutoff of -40 kJ/mol, which is about 40-fold over random screening (2.06%). Limited results suggest that presenting more poses of a single molecule to the scoring functions could deteriorate their enrichment factors. A series of promising scaffolds with favorable binding scores was retrieved from LeadQuest. Consensus scoring by pair-wise intersection failed to enrich the hit rate yielded by single scorings (i.e. FlexX). We note that reported successes of consensus scoring in hit rate enrichment could be artificial because their comparisons were based on a selected subset of single scoring and a markedly reduced subset of double or triple scoring. The findings presented in this report are based upon a single biological system and support further studies.

• Multiple active site corrections for docking and virtual screening.
Vigers, Guy P A and Rizzi, James P
Journal of medicinal chemistry, 2004, 47(1), 80-89
PMID: 14695822     doi: 10.1021/jm030161o

Several docking programs are now available that can reproduce the bound conformation of a ligand in an active site, for a wide variety of experimentally determined complexes. However, these programs generally perform less well at ranking multiple possible ligands in one site. Since accurate identification of potential ligands is a prerequisite for many aspects of structure-based drug design, this is a serious limitation. We have tested the ability of two docking programs, FlexX and Gold, to match ligands and active sites for multiple complexes. We show that none of the docking scores from either program are able to match consistently ligands and active sites in our tests. We propose a simple statistical correction, the multiple active site correction (MASC), which greatly ameliorates this problem. We have also tested the correction method against an extended set of 63 cocrystals and in a virtual screening experiment. In all cases, MASC significantly improves the results of the docking experiments.

• Glide: A new approach for rapid, accurate docking and scoring. 1. Method and assessment of docking accuracy
Friesner, RA and Banks, JL and Murphy, RB and Halgren, TA and Klicic, JJ and Mainz, DT and Repasky, MP and Knoll, EH and Shelley, M and Perry, JK and Shaw, DE and Francis, P and Shenkin, PS
Journal of medicinal chemistry, 2004, 47(7), 1739-1749
PMID: 15027865     doi: 10.1021/jm0306430

Unlike other methods for docking ligands to the rigid 3D structure of a known protein receptor, Glide approximates a complete systematic search of the conformational, orientational, and positional space of the docked ligand. In this search, an initial rough positioning and scoring phase that dramatically narrows the search space is followed by torsionally flexible energy optimization on an OPLS-AA nonbonded potential grid for a few hundred surviving candidate poses. The very best candidates are further refined via a Monte Carlo sampling of pose conformation; in some cases, this is crucial to obtaining an accurate docked pose. Selection of the best docked pose uses a model energy function that combines empirical and force-field-based terms. Docking accuracy is assessed by redocking ligands from 282 cocrystallized PDB complexes starting from conformationally optimized ligand geometries that bear no memory of the correctly docked pose. Errors in geometry for the top-ranked pose are less than 1 Angstrom in nearly half of the cases and are greater than 2 Angstrom in only about one-third of them. Comparisons to published data on rms deviations show that Glide is nearly twice as accurate as GOLD and more than twice as accurate as FlexX for ligands having up to 20 rotatable bonds. Glide is also found to be more accurate than the recently described Surflex method.

• Assessing scoring functions for protein-ligand interactions.
Ferrara, Philippe and Gohlke, Holger and Price, Daniel J and Klebe, Gerhard and Brooks, Charles L
Journal of medicinal chemistry, 2004, 47(12), 3032-3047
PMID: 15163185     doi: 10.1021/jm030489h

An assessment of nine scoring functions commonly applied in docking using a set of 189 protein-ligand complexes is presented. The scoring functions include the CHARMm potential, the scoring function DrugScore, the scoring function used in AutoDock, the three scoring functions implemented in DOCK, as well as three scoring functions implemented in the CScore module in SYBYL (PMF, Gold, ChemScore). We evaluated the abilities of these scoring functions to recognize near-native configurations among a set of decoys and to rank binding affinities. Binding site decoys were generated by molecular dynamics with restraints. To investigate whether the scoring functions can also be applied for binding site detection, decoys on the protein surface were generated. The influence of the assignment of protonation states was probed by either assigning "standard" protonation states to binding site residues or adjusting protonation states according to experimental evidence. The role of solvation models in conjunction with CHARMm was explored in detail. These include a distance-dependent dielectric function, a generalized Born model, and the Poisson equation. We evaluated the effect of using a rigid receptor on the outcome of docking by generating all-pairs decoys ("cross-decoys") for six trypsin and seven HIV-1 protease complexes. The scoring functions perform well to discriminate near-native from misdocked conformations, with CHARMm, DOCK-energy, DrugScore, ChemScore, and AutoDock yielding recognition rates of around 80%. Significant degradation in performance is observed in going from decoy to cross-decoy recognition for CHARMm in the case of HIV-1 protease, whereas DrugScore and ChemScore, as well as CHARMm in the case of trypsin, show only small deterioration. In contrast, the prediction of binding affinities remains problematic for all of the scoring functions. ChemScore gives the highest correlation value with R(2)

## 2003

• Molecular recognition and docking algorithms.
Brooijmans, Natasja and Kuntz, Irwin D
Annual review of biophysics and biomolecular structure, 2003, 32, 335-373
PMID: 12574069     doi: 10.1146/annurev.biophys.32.110601.142532

Molecular docking is an invaluable tool in modern drug discovery. This review focuses on methodological developments relevant to the field of molecular docking. The forces important in molecular recognition are reviewed and followed by a discussion of how different scoring functions account for these forces. More recent applications of computational chemistry tools involve library design and database screening. Last, we summarize several critical methodological issues that must be addressed in future developments.

• Comparative evaluation of 11 scoring functions for molecular docking.
Wang, Renxiao and Lu, Yipin and Wang, Shaomeng
Journal of medicinal chemistry, 2003, 46(12), 2287-2303
PMID: 12773034     doi: 10.1021/jm0203783

Eleven popular scoring functions have been tested on 100 protein-ligand complexes to evaluate their abilities to reproduce experimentally determined structures and binding affinities. They include four scoring functions implemented in the LigFit module in Cerius2 (LigScore, PLP, PMF, and LUDI), four scoring functions implemented in the CScore module in SYBYL (F-Score, G-Score, D-Score, and ChemScore), the scoring function implemented in the AutoDock program, and two stand-alone scoring functions (DrugScore and X-Score). These scoring functions are not tested in the context of a particular docking program. Instead, conformational sampling and scoring are separated into two consecutive steps. First, an exhaustive conformational sampling is performed by using the AutoDock program to generate an ensemble of docked conformations for each ligand molecule. This conformational ensemble is required to cover the entire conformational space as much as possible rather than to focus on a few energy minima. Then, each scoring function is applied to score this conformational ensemble to see if it can identify the experimentally observed conformation from all of the other decoys. Among all of the scoring functions under test, six of them, i.e., PLP, F-Score, LigScore, DrugScore, LUDI, and X-Score, yield success rates higher than the AutoDock scoring function. The success rates of these six scoring functions range from 66% to 76% if using root-mean-square deviation < or

• FDS: flexible ligand and receptor docking with a continuum solvent model and soft-core energy function.
Taylor, Richard D and Jewsbury, Philip J and Essex, Jonathan W
Journal of computational chemistry, 2003, 24(13), 1637-1656
PMID: 12926007     doi: 10.1002/jcc.10295

The docking of flexible small molecule ligands to large flexible protein targets is addressed in this article using a two-stage simulation-based method. The methodology presented is a hybrid approach where the first component is a dock of the ligand to the protein binding site, based on deriving sets of simultaneously satisfied intermolecular hydrogen bonds using graph theory and a recursive distance geometry algorithm. The output structures are reduced in number by cluster analysis based on distance similarities. These structures are submitted to a modified Monte Carlo algorithm using the AMBER-AA molecular mechanics force field with the Generalized Born/Surface Area (GB/SA) continuum model. This solvent model is not only less expensive than an explicit representation, but also yields increased sampling. Sampling is also increased using a rotamer library to direct some of the protein side-chain movements along with large dihedral moves. Finally, a softening function for the nonbonded force field terms is used, enabling the potential energy function to be slowly turned on throughout the course of the simulation. The docking procedure is optimized, and the results are presented for a single complex of the arabinose binding protein. It was found that for a rigid receptor model, the X-ray binding geometry was reproduced and uniquely identified based on the associated potential energy. However, when side-chain flexibility was included, although the X-ray structure was identified, it was one of three possible binding geometries that were energetically indistinguishable. These results suggest that on relaxing the constraint on receptor flexibility, the docking energy hypersurface changes from being funnel-like to rugged. A further 14 complexes were then examined using the optimized protocol. For each complex the docking methodology was tested for a fully flexible ligand, both with and without protein side-chain flexibility. For the rigid protein docking, 13 out of the 15 test cases were able to find the experimental binding mode; this number was reduced to 11 for the flexible protein docking. However, of these 11, in the majority of cases the experimental binding mode was not uniquely identified, but was present in a cluster of low energy structures that were energetically indistinguishable. These results not only support the presence of a rugged docking energy hypersurface, but also suggest that it may be necessary to consider the possibility of more than one binding conformation during ligand optimization.

• Gaussian docking functions.
McGann, Mark R and Almond, Harold R and Nicholls, Anthony and Grant, J Andrew and Brown, Frank K
Biopolymers, 2003, 68(1), 76-90
PMID: 12579581     doi: 10.1002/bip.10207

A shape-based Gaussian docking function is constructed which uses Gaussian functions to represent the shapes of individual atoms. A set of 20 trypsin ligand-protein complexes are drawn from the Protein Data Bank (PDB), the ligands are separated from the proteins, and then are docked back into the active sites using numerical optimization of this function. It is found that by employing this docking function, quasi-Newton optimization is capable of moving ligands great distances [on average 7 A root mean square distance (RMSD)] to locate the correctly docked structure. It is also found that a ligand drawn from one PDB file can be docked into a trypsin structure drawn from any of the trypsin PDB files. This implies that this scoring function is not limited to more accurate x-ray structures, as is the case for many of the conventional docking methods, but could be extended to homology models.

• Improved protein-ligand docking using GOLD.
Verdonk, Marcel L and Cole, Jason C and Hartshorn, Michael J and Murray, Christopher W and Taylor, Richard D
Proteins, 2003, 52(4), 609-623
PMID: 12910460     doi: 10.1002/prot.10465

The Chemscore function was implemented as a scoring function for the protein-ligand docking program GOLD, and its performance compared to the original Goldscore function and two consensus docking protocols, "Goldscore-CS" and "Chemscore-GS," in terms of docking accuracy, prediction of binding affinities, and speed. In the "Goldscore-CS" protocol, dockings produced with the Goldscore function are scored and ranked with the Chemscore function; in the "Chemscore-GS" protocol, dockings produced with the Chemscore function are scored and ranked with the Goldscore function. Comparisons were made for a "clean" set of 224 protein-ligand complexes, and for two subsets of this set, one for which the ligands are "drug-like," the other for which they are "fragment-like." For "drug-like" and "fragment-like" ligands, the docking accuracies obtained with Chemscore and Goldscore functions are similar. For larger ligands, Goldscore gives superior results. Docking with the Chemscore function is up to three times faster than docking with the Goldscore function. Both combined docking protocols give significant improvements in docking accuracy over the use of the Goldscore or Chemscore function alone. "Goldscore-CS" gives success rates of up to 81% (top-ranked GOLD solution within 2.0 A of the experimental binding mode) for the "clean list," but at the cost of long search times. For most virtual screening applications, "Chemscore-GS" seems optimal; search settings that give docking speeds of around 0.25-1.3 min/compound have success rates of about 78% for "drug-like" compounds and 85% for "fragment-like" compounds. In terms of producing binding energy estimates, the Goldscore function appears to perform better than the Chemscore function and the two consensus protocols, particularly for faster search settings. Even at docking speeds of around 1-2 min/compound, the Goldscore function predicts binding energies with a standard deviation of approximately 10.5 kJ/mol.

## 2002

• Approaches to the description and prediction of the binding affinity of small-molecule ligands to macromolecular receptors.
Gohlke, Holger and Klebe, Gerhard
Angewandte Chemie (International ed. in English), 2002, 41(15), 2644-2676
PMID: 12203463     doi: 10.1002/1521-3773(20020802)41:15<2644::AID-ANIE2644>3.0.CO;2-O

The influence of a xenobiotic compound on an organism is usually summarized by the expression biological activity. If a controlled, therapeutically relevant, and regulatory action is observed the compound has potential as a drug, otherwise its toxicity on the biological system is of interest. However, what do we understand by the biological activity? In principle, the overall effect on an organism has to be considered. However, because of the complexity of the interrelated processes involved, as a simplification primarily the "main action" on the organism is taken into consideration. On the molecular level, biological activity corresponds to the binding of a (low-molecular weight) compound to a macromolecular receptor, usually a protein. Enzymatic reactions or signal-transduction cascades are thereby influenced with respect to their function for the organism. We regard this binding as a process under equilibrium conditions; thus, binding can be described as an association or dissociation process. Accordingly, biological activity is expressed as the affinity of both partners for each other, as a thermodynamic equilibrium quantity. How well do we understand these terms and how well are they theoretically predictable today? The holy grail of rational drug design is the prediction of the biological activity of a compound. The processes involving ligand binding are extremely complicated, both ligand and protein are flexible molecules, and the energy inventory between the bound and unbound states must be considered in aqueous solution. How sophisticated and reliable are our experimental approaches to obtaining the necessary insight? The present review summarizes our current understanding of the binding affinity of a small-molecule ligand to a protein. Both theoretical and empirical approaches for predicting binding affinity, starting from the three-dimensional structure of a protein-ligand complex, will be described and compared. Experimental methods, primarily microcalorimetry, will be discussed. As a perspective, our own knowledge-based approach towards affinity prediction and experimental data on factorizing binding contributions to protein-ligand binding will be presented.

• Further development and validation of empirical scoring functions for structure-based binding affinity prediction
Wang, R and Lai, L
Journal of computer-aided molecular design, 2002, 16, 11-26
PMID: 12197663

New empirical scoring functions have been developed to estimate the binding affinity of a given protein-ligand complex with known three-dimensional structure. These scoring functions include terms accounting for van der Waals interaction, hydrogen bonding, deformation penalty, and hydrophobic effect. A special feature is that three different algorithms have been implemented to calculate the hydrophobic effect term, which results in three parallel scoring functions. All three scoring functions are calibrated through multivariate regression analysis of a set of 200 protein-ligand complexes and they reproduce the binding free energies of the entire training set with standard deviations of 2.2 kcal/mol, 2.1 kcal/mol, and 2.0 kcal/mol, respectively. These three scoring functions are further combined into a consensus scoring function, X-CSCORE. When tested on an independent set of 30 protein-ligand complexes, X-CSCORE is able to predict their binding free energies with a standard deviation of 2.2 kcal/mol. The potential application of X-CSCORE to molecular docking is also investigated. Our results show that this consensus scoring function improves the docking accuracy considerably when compared to the conventional force field computation used for molecular docking.

• Simple, intuitive calculations of free energy of binding for protein-ligand complexes. 1. Models without explicit constrained water.
Cozzini, Pietro and Fornabaio, Micaela and Marabotti, Anna and Abraham, Donald J and Kellogg, Glen E and Mozzarelli, Andrea
Journal of medicinal chemistry, 2002, 45(12), 2469-2483
PMID: 12036355

The prediction of the binding affinity between a protein and ligands is one of the most challenging issues for computational biochemistry and drug discovery. While the enthalpic contribution to binding is routinely available with molecular mechanics methods, the entropic contribution is more difficult to estimate. We describe and apply a relatively simple and intuitive calculation procedure for estimating the free energy of binding for 53 protein-ligand complexes formed by 17 proteins of known three-dimensional structure and characterized by different active site polarity. HINT, a software model based on experimental LogP(o/w) values for small organic molecules, was used to evaluate and score all atom-atom hydropathic interactions between the protein and the ligands. These total scores (H(TOTAL)), which have been previously shown to correlate with DeltaG(interaction) for protein-protein interactions, correlate with DeltaG(binding) for protein-ligand complexes in the present study with a standard error of +/-2.6 kcal mol(-1) from the equation DeltaG(binding)

• Consensus scoring for ligand/protein interactions.
Clark, Robert D and Strizhev, Alexander and Leonard, Joseph M and Blake, James F and Matthew, James B
Journal of molecular graphics & modelling, 2002, 20(4), 281-295
PMID: 11858637

Several different functions have been put forward for evaluating the energetics of ligand binding to proteins. Those employed in the DOCK, GOLD and FlexX docking programs have been especially widely used, particularly in connection with virtual high-throughput screening (vHTS) projects. Until recently, such evaluation functions were usually considered only in conjunction with the docking programs that relied on them. In such studies, the evaluation function in question actually fills two distinct roles: it serves as the objective function being optimized (fitness function), but is also the scoring function used to compare the candidate docking configurations generated by the program. We have used descriptions available in the open literature to create free-standing scoring functions based on those used in DOCK and GOLD, and have implemented the more recently formulated PMF [J. Med. Chem. 42 (1999) 791] scoring function as well. The performance of these functions was examined individually for each of several data sets for which both crystal structures and affinities are available, as was the performance of the FlexX scoring function. Various ways of combining individual scores into a consensus score (CScore) were also considered. The individual and consensus scores were also used to try to pick out configurations most similar to those found in crystal structures from among a set of candidate configurations produced by FlexX docking runs. We find that the reliability and interpretability of results can be improved by combining results from all four functions into a CScore.

## 2001

• Evaluation of docking functions for protein-ligand docking
Perez, C and Ortiz, AR
Journal of medicinal chemistry, 2001, 44(23), 3768-3785
doi: 10.1021/jm010141r

Docking functions are believed to be the essential component of docking algorithms. Both physically and statistically based functions have been proposed, but there is no consensus about their relative performances. Here, we propose an evaluation approach based on exhaustive enumeration of all possible docking solutions obtained with a discretized description of a rigid docking process. We apply the approach to study both molecular mechanics and statistical potentials. It is found that the statistical potential evaluated is less effective than the AMBER molecular mechanics function to provide an accurate description of the docking process when the exact experimental coordinates are used. However, when coordinates of crystal structures obtained with analogous ligands are used, similar performances are obtained in both cases. Possible reasons for the successes and failures of both docking schemes have been uncovered using linear discriminant analysis, on the basis of a set of physicochemical descriptors capturing the main physical effects at play during protein-ligand docking. In both types of potentials steric effects appear critical to obtain a successful docking. Our results also indicate that neglecting desolvation effects and the explicit treatment of hydrogen bonds are the main source of the failures observed with the molecular mechanics potential. On the other hand, detailed consideration of steric interactions, with a careful treatment of dispersive forces, seems to be needed when using statistical potentials derived from a structural database. The possibility of filtering combinatorial libraries in order to maximize the probability of correct docking is discussed.

• Detailed analysis of scoring functions for virtual screening.
Stahl, M and Rarey, M
Journal of medicinal chemistry, 2001, 44(7), 1035-1042
PMID: 11297450

We present a comprehensive study of the performance of fast scoring functions for library docking using the program FlexX as the docking engine. Four scoring functions, among them two recently developed knowledge-based potentials, are evaluated on seven target proteins whose binding sites represent a wide range of size, form, and polarity. The results of these calculations give valuable insight into strengths and weaknesses of current scoring functions. Furthermore, it is shown that a well-chosen combination of two of the tested scoring functions leads to a new, robust scoring scheme with superior performance in virtual screening.

## 2000

• Knowledge-based scoring function to predict protein-ligand interactions.
Gohlke, H and Hendlich, M and Klebe, G
Journal of molecular biology, 2000, 295(2), 337-356
PMID: 10623530     doi: 10.1006/jmbi.1999.3371

The development and validation of a new knowledge-based scoring function (DrugScore) to describe the binding geometry of ligands in proteins is presented. It discriminates efficiently between well-docked ligand binding modes (root-mean-square deviation <2.0 A with respect to a crystallographically determined reference complex) and those largely deviating from the native structure, e.g. generated by computer docking programs. Structural information is extracted from crystallographically determined protein-ligand complexes using ReLiBase and converted into distance-dependent pair-preferences and solvent-accessible surface (SAS) dependent singlet preferences for protein and ligand atoms. Definition of an appropriate reference state and accounting for inaccuracies inherently present in experimental data is required to achieve good predictive power. The sum of the pair preferences and the singlet preferences is calculated based on the 3D structure of protein-ligand binding modes generated by docking tools. For two test sets of 91 and 68 protein-ligand complexes, taken from the Protein Data Bank (PDB), the calculated score recognizes poses generated by FlexX deviating <2 A from the crystal structure on rank 1 in three quarters of all possible cases. Compared to FlexX, this is a substantial improvement. For ligand geometries generated by DOCK, DrugScore is superior to the "chemical scoring" implemented into this tool, while comparable results are obtained using the "energy scoring" in DOCK. None of the presently known scoring functions achieves comparable power to extract binding modes in agreement with experiment. It is fast to compute, regards implicitly solvation and entropy contributions and produces correctly the geometry of directional interactions. Small deviations in the 3D structure are tolerated and, since only contacts to non-hydrogen atoms are regarded, it is independent from assumptions of protonation states.

## 1999

• A general and fast scoring function for protein-ligand interactions: a simplified potential approach.
Muegge, I and Martin, Y C
Journal of medicinal chemistry, 1999, 42(5), 791-804
PMID: 10072678     doi: 10.1021/jm980536j

A fast, simplified potential-based approach is presented that estimates the protein-ligand binding affinity based on the given 3D structure of a protein-ligand complex. This general, knowledge-based approach exploits structural information of known protein-ligand complexes extracted from the Brookhaven Protein Data Bank and converts it into distance-dependent Helmholtz free interaction energies of protein-ligand atom pairs (potentials of mean force, PMF). The definition of an appropriate reference state and the introduction of a correction term accounting for the volume taken by the ligand were found to be crucial for deriving the relevant interaction potentials that treat solvation and entropic contributions implicitly. A significant correlation between experimental binding affinities and computed score was found for sets of diverse protein-ligand complexes and for sets of different ligands bound to the same target. For 77 protein-ligand complexes taken from the Brookhaven Protein Data Bank, the calculated score showed a standard deviation from observed binding affinities of 1.8 log Ki units and an R2 value of 0.61. The best results were obtained for the subset of 16 serine protease complexes with a standard deviation of 1.0 log Ki unit and an R2 value of 0.86. A set of 33 inhibitors modeled into a crystal structure of HIV-1 protease yielded a standard deviation of 0.8 log Ki units from measured inhibition constants and an R2 value of 0.74. In contrast to empirical scoring functions that show similar or sometimes better correlation with observed binding affinities, our method does not involve deriving specific parameters that fit the observed binding affinities of protein-ligand complexes of a given training set. We compared the performance of the PMF score, Böhm's score (LUDI), and the SMOG score for eight different test sets of protein-ligand complexes. It was found that for the majority of test sets the PMF score performs best. The strength of the new approach presented here lies in its generality as no knowledge about measured binding affinities is needed to derive atomic interaction potentials. The use of the new scoring function in docking studies is outlined.

• BLEEP - potential of mean force describing protein-ligand interactions: II. Calculation of binding energies and comparison with experimental data
Alex, A and Forster, MJ and Thornton, JM
Journal of computational chemistry, 1999, 20(11), 1177-1185

We have developed BLEEP\v Z}biomolecular ligand energy evaluation protocol., an atomic level potential of mean force\v Z}PMF. describing protein􏱌ligand interactions. Here, we present four tests designed to assess different attributes of BLEEP. Calculating the energy of a small hydrogen-bonded complex allows us to compare BLEEP's description of this system with a quantum-chemical description. The results suggest that BLEEP gives an adequate description of hydrogen bonding. A study of the relative energies of various heparin binding geometries for human basic fibroblast growth factor\v Z}bFGF. demonstrates that BLEEP performs excellently in identifying low-energy binding modes from decoy conformations for a given protein􏱌ligand complex. We also calculate binding energies for a set of 90 protein􏱌ligand complexes, obtaining a correlation coefficient of 0.74 when compared with experiment. This shows that BLEEP can perform well in the difficult area of ranking the interaction energies of diverse complexes. We also study a set of nine serine proteinase􏱌inhibitor complexes; BLEEP's good performance here illustrates its ability to determine the relative energies of a series of similar complexes. We find that a protocol for incorporating solvation does not improve correlation with experiment.

• Consensus scoring: A method for obtaining improved hit rates from docking databases of three-dimensional structures into proteins.
Charifson, P S and Corkery, J J and Murcko, M A and Walters, W P
Journal of medicinal chemistry, 1999, 42(25), 5100-5109
PMID: 10602695

We present the results of an extensive computational study in which we show that combining scoring functions in an intersection-based consensus approach results in an enhancement in the ability to discriminate between active and inactive enzyme inhibitors. This is illustrated in the context of docking collections of three-dimensional structures into three different enzymes of pharmaceutical interest: p38 MAP kinase, inosine monophosphate dehydrogenase, and HIV protease. An analysis of two different docking methods and thirteen scoring functions provides insights into which functions perform well, both singly and in combination. Our data shows that consensus scoring further provides a dramatic reduction in the number of false positives identified by individual scoring functions, thus leading to a significant enhancement in hit-rates.

• Exhaustive docking of molecular fragments with electrostatic solvation.
Majeux, N and Scarsi, M and Apostolakis, J and Ehrhardt, C and Caflisch, A
Proteins, 1999, 37(1), 88-105
PMID: 10451553

A new method is presented for docking molecular fragments to a rigid protein with evaluation of the binding energy. Polar fragments are docked with at least one hydrogen bond with the protein while apolar fragments are positioned in the hydrophobic pockets. The electrostatic contribution to the binding energy, which consists of screened intermolecular energy and protein and fragment desolvation terms, is evaluated efficiently by a numerical approach based on the continuum dielectric approximation. The latter is also used to predetermine the hydrophobic pockets of the protein by rolling a low dielectric sphere over the protein surface and calculating the electrostatic desolvation of the protein and van der Waals interaction energy. The method was implemented in the program SEED (solvation energy for exhaustive docking). The SEED continuum electrostatic approach has been successfully validated by a comparison with finite difference solutions of the Poisson equation for more than 2,500 complexes of small molecules with thrombin and the monomer of HIV-1 aspartic proteinase. The fragments docked by SEED in the active site of thrombin reproduce the structural features of the interaction patterns between known inhibitors and thrombin. Moreover, the combinatorial connection of these fragments yields a number of compounds that are very similar to potent inhibitors of thrombin. Proteins 1999;37:88-105.

## 1998

• Automated docking using a Lamarckian genetic algorithm and an empirical binding free energy function
Morris, GM and Goodsell, DS and Halliday, RS
Journal of computational chemistry, 1998, 19(14), 1639-1662

A novel and robust automated docking method that predicts the bound conformations of flexible ligands to macromolecular targets has been developed and tested, in combination with a new scoring function that estimates the free energy change upon binding. Interestingly, this method applies a Lamarckian model of genetics, in which environmental adaptations of an individual's phenotype are reverse transcribed into its genotype and become heritable traits\v Z}sic.. We consider three search methods, Monte Carlo simulated annealing, a traditional genetic algorithm, and the Lamarckian genetic algorithm, and compare their performance in dockings of seven protein􏱌ligand test systems having known three-dimensional structure. We show that both the traditional and Lamarckian genetic algorithms can handle ligands with more degrees of freedom than the simulated annealing method used in earlier versions of AUTODOCK, and that the Lamarckian genetic algorithm is the most efficient, reliable, and successful of the three. The empirical free energy function was calibrated using a set of 30 structurally known protein􏱌ligand complexes with experimentally determined binding constants. Linear regression analysis of the observed binding constants in terms of a wide variety of structure-derived molecular properties was performed. The final model had a residual standard error of 9.11 kJ mol􏳡1\v Z}2.177 kcal mol􏳡1 . and was chosen as the new energy

• Screening a peptidyl database for potential ligands to proteins with side-chain flexibility.
Schnecke, V and Swanson, C A and Getzoff, E D and Tainer, J A and Kuhn, L A
Proteins, 1998, 33(1), 74-87
PMID: 9741846

The three key challenges addressed in our development of SPECITOPE, a tool for screening large structural databases for potential ligands to a protein, are to eliminate infeasible candidates early in the search, incorporate ligand and protein side-chain flexibility upon docking, and provide an appropriate rank for potential new ligands. The protein ligand-binding site is modeled by a shell of surface atoms and by hydrogen-bonding template points for the ligand to match, conferring specificity to the interaction. SPECITOPE combinatorially matches all hydrogen-bond donors and acceptors of the screened molecules to the template points. By eliminating molecules that cannot match distance or hydrogen-bond constraints, the transformation of potential docking candidates into the ligand-binding site and the shape and hydrophobic complementarity evaluations are only required for a small subset of the database. SPECITOPE screens 140,000 peptide fragments in about an hour and has identified and docked known inhibitors and potential new ligands to the free structures of four distinct targets: a serine protease, a DNA repair enzyme, an aspartic proteinase, and a glycosyltransferase. For all four, protein side-chain rotations were critical for successful docking, emphasizing the importance of inducible complementarity for accurately modeling ligand interactions. SPECITOPE has a range of potential applications for understanding and engineering protein recognition, from inhibitor and linker design to protein docking and macromolecular assembly.

• Prediction of binding constants of protein ligands: a fast method for the prioritization of hits obtained from de novo design or 3D database search programs.
Böhm, H J
Journal of computer-aided molecular design, 1998, 12(4), 309-323
PMID: 9777490

A dataset of 82 protein-ligand complexes of known 3D structure and binding constant Ki was analysed to elucidate the important factors that determine the strength of protein-ligand interactions. The following parameters were investigated: the number and geometry of hydrogen bonds and ionic interactions between the protein and the ligand, the size of the lipophilic contact surface, the flexibility of the ligand, the electrostatic potential in the binding site, water molecules in the binding site, cavities along the protein-ligand interface and specific interactions between aromatic rings. Based on these parameters, a new empirical scoring function is presented that estimates the free energy of binding for a protein-ligand complex of known 3D structure. The function distinguishes between buried and solvent accessible hydrogen bonds. It tolerates deviations in the hydrogen bond geometry of up to 0.25 A in the length and up to 30 degrees in the hydrogen bond angle without penalizing the score. The new energy function reproduces the binding constants (ranging from 3.7 x 10(-2) M to 1 x 10(-14) M, corresponding to binding energies between -8 and -80 kJ/mol) of the dataset with a standard deviation of 7.3 kJ/mol corresponding to 1.3 orders of magnitude in binding affinity. The function can be evaluated very fast and is therefore also suitable for the application in a 3D database search or de novo ligand design program such as LUDI. The physical significance of the individual contributions is discussed.

• Empirical scoring functions. II. The testing of an empirical scoring function for the prediction of ligand-receptor binding affinities and the use of Bayesian regression to improve the quality of the model.
Murray, C W and Auton, T R and Eldridge, M D
Journal of computer-aided molecular design, 1998, 12(5), 503-519
PMID: 9834910

This paper tests the performance of a simple empirical scoring function on a set of candidate designs produced by a de novo design package. The scoring function calculates approximate ligand-receptor binding affinities given a putative binding geometry. To our knowledge this is the first substantial test of an empirical scoring function of this type on a set of molecular designs which were then subsequently synthesised and assayed. The performance illustrates that the methods used to construct the scoring function and the reliance on plausible, yet potentially false, binding modes can lead to significant over-prediction of binding affinity in bad cases. This is anticipated on theoretical grounds and provides caveats on the reliance which can be placed when using the scoring function as a screen in the choice of molecular designs. To improve the predictability of the scoring function and to understand experimental results, it is important to perform subsequent Quantitative Structure-Activity Relationship (QSAR) studies. In this paper, Bayesian regression is performed to improve the predictability of the scoring function in the light of the assay results. Bayesian regression provides a rigorous mathematical framework for the incorporation of prior information, in this case information from the original training set, into a regression on the assay results of the candidate molecular designs. The results indicate that Bayesian regression is a useful and practical technique when relevant prior knowledge is available and that the constraints embodied in the prior information can be used to improve the robustness and accuracy of regression models. We believe this to be the first application of Bayesian regression to QSAR analysis in chemistry.

• Assessing energy functions for flexible docking
Vieth, M and Hirst, JD and Kolinski, A and Brooks, CL
Journal of computational chemistry, 1998, 19(14), 1612-1622

A good docking algorithm requires an energy function that is selective, in that it clearly differentiates correctly docked structures from misdocked ones, and that is efficient, meaning that a correctly docked structure can be identified quickly. We assess the selectivity and efficiency of a broad spectrum of energy functions, derived from systematic modifications of the CHARMM param19/toph19 energy function, in particular, we examine the effects of the dielectric constant, the solvation model, the scaling of surface charges, reduction of van der Waals repulsion, and nonbonded cutoffs. Based on an assessment of the energy functions for the docking of five different Ligand-receptor complexes, we find that selective energy functions include a variety of distance-dependent dielectric models together with truncation of the nonbonded interactions at 8 Angstrom. We evaluate the docking efficiency, the mean number of docked structures per unit of time, of the more selective energy functions, using a simulated annealing molecular dynamics protocol. The largest improvements in efficiency come from a reduction of van der Waals repulsion and a reduction of surface charges. We note that the most selective potential is quite inefficient, although a hierarchical approach can be employed to take advantage of both selective and efficient energy functions. (C) 1998 John Wiley & Sons, Inc.

## 1997

• SMoG:  de Novo Design Method Based on Simple, Fast, and Accurate Free Energy Estimates. 2. Case Studies in Molecular Design
Robert S DeWitte and Alexey V Ishchenko, and and Shakhnovich, Eugene I
Journal of the American Chemical Society, 1997, 119(20), 4608-4617

In this paper, we summarize three ligand design studies performed using the program SMoG, which was developed in our lab. The aim of this presentation is to communicate through examples the potential of this method:  the richness of the molecules that can be developed and the ease with which they are found. In particular, we present suggestions for ligands to Src SH3 domain (specificity pocket and LP site) and CD4.

• Empirical scoring functions: I. The development of a fast empirical scoring function to estimate the binding affinity of ligands in receptor complexes.
Eldridge, M D and Murray, C W and Auton, T R and Paolini, G V and Mee, R P
Journal of computer-aided molecular design, 1997, 11(5), 425-445
PMID: 9385547

This paper describes the development of a simple empirical scoring function designed to estimate the free energy of binding for a protein-ligand complex when the 3D structure of the complex is known or can be approximated. The function uses simple contact terms to estimate lipophilic and metal-ligand binding contributions, a simple explicit form for hydrogen bonds and a term which penalises flexibility. The coefficients of each term are obtained using a regression based on 82 ligand-receptor complexes for which the binding affinity is known. The function reproduces the binding affinity of the complexes with a cross-validated error of 8.68 kJ/mol. Tests on internal consistency indicate that the coefficients obtained are stable to changes in the composition of the training set. The function is also tested on two test sets containing a further 20 and 10 complexes, respectively. The deficiencies of this type of function are discussed and it is compared to approaches by other workers.

• QXP: Powerful, rapid computer algorithms for structure-based drug design
McMartin, C and Bohacek, RS
Journal of computer-aided molecular design, 1997, 11(4), 333-344
PMID: 9334900

New methods for docking, template fitting and building pseudo-receptors are described. Full conformational searches are carried out for flexible cyclic and acyclic molecules. QXP (quick explore) search algorithms are derived from the method of Monte Carlo perturbation with energy minimization in Cartesian space. An additional fast search step is introduced between the initial perturbation and energy minimization. The fast search produces approximate low-energy structures, which are likely to minimize to a low energy. For template fitting, QXP uses a superposition force field which automatically assigns short-range attractive forces to similar atoms in different molecules. The docking algorithms were evaluated using X-ray data for 12 protein-ligand complexes. The ligands had up to 24 rotatable bonds and ranged from highly polar to mostly nonpolar. Docking searches of the randomly disordered ligands gave rms differences between the lowest energy docked structure and the energy-minimized X-ray structure, of less than 0.76 Angstrom for 10 of the ligands. For all the ligands, the rms difference between the energy-minimized X-ray structure and the closest docked structure was less than 0.4 Angstrom, when parts of one of the molecules which are in the solvent were excluded from the rms calculation. Template fitting was tested using four ACE inhibitors. Three ACE templates have been previously published. A single run using QXP generated a series of templates which contained examples of each of the three. A pseudo-receptor, complementary to an ACE template, was built out of small molecules, such as pyrrole, cyclo-pentanone and propane. When individually energy minimized in the pseudo-receptor, each of the four ACE inhibitors moved with an rms of less than 0.25 Angstrom. After random perturbation, the inhibitors were docked into the pseudo-receptor. Each lowest energy docked structure matched the energy-minimized geometry with an rms of less than 0.08 Angstrom. Thus, the pseudo-receptor shows steric and chemical complementarity to all four molecules. The QXP program is reliable, easy to use and sufficiently rapid for routine application in structure-based drug design.

• Development and validation of a genetic algorithm for flexible docking.
Jones, G and Willett, P and Glen, R C and Leach, A R and Taylor, R
Journal of molecular biology, 1997, 267(3), 727-748
PMID: 9126849     doi: 10.1006/jmbi.1996.0897

Prediction of small molecule binding modes to macromolecules of known three-dimensional structure is a problem of paramount importance in rational drug design (the "docking" problem). We report the development and validation of the program GOLD (Genetic Optimisation for Ligand Docking). GOLD is an automated ligand docking program that uses a genetic algorithm to explore the full range of ligand conformational flexibility with partial flexibility of the protein, and satisfies the fundamental requirement that the ligand must displace loosely bound water on binding. Numerous enhancements and modifications have been applied to the original technique resulting in a substantial increase in the reliability and the applicability of the algorithm. The advanced algorithm has been tested on a dataset of 100 complexes extracted from the Brookhaven Protein DataBank. When used to dock the ligand back into the binding site, GOLD achieved a 71% success rate in identifying the experimental binding mode.

## 1996

• SMoG: de Novo Design Method Based on Simple, Fast, and Accurate Free Energy Estimates. 1. Methodology and Supporting Evidence
DeWitte, Robert S and Shakhnovich, Eugene I
Journal of the American Chemical Society, 1996, 118(47), 11733-11744
doi: 10.1021/ja960751u

In this paper, we present SMoG (Small Molecule Growth), a novel, straightforward method for de novo lead design and the evidence for its effectiveness. It is based on a simple model for ligand-protein interactions and a scoring that is directly related to the free energy through a knowledge-based potential. A large number of structures are examined by an efficient metropolis Monte Carlo molecular growth algorithm that generates molecules through the adjoining of functional groups directly in the binding region. Thus SMoG is a method that is able to rank a large number of potential compounds according to binding free energy in a short time. In this sense, SMoG represents a step toward an ideal computational tool for ligand design.

• VALIDATE: A New Method for the Receptor-Based Prediction of Binding Affinities of Novel Ligands
Head, Richard D and Smythe, Mark L and Oprea, Tudor I and Waller, Chris L and Green, Stuart M and Marshall, Garland R.
Journal of the American Chemical Society, 1996, 118(16), 3959-3969
doi: 10.1021/ja9539002

VALIDATE is a hybrid approach to predict the binding affinity of novel ligands for receptors of known three-dimensional structure. This approach calculates physicochemical properties of the ligand and the receptor- ligand complex to estimate the free energy of binding. The enthalpy of binding is calculated by molecular mechanics while properties such as complementary hydrophobic surface area are used to estimate the entropy of binding through heuristics. A diverse training set of 51 crystalline complexes was assembled, and their relevant physicochemical properties were computed. These properties were analyzed by partial least squares (PLS) statistics, or neural network analysis (SONNIC), to generate models for the general prediction of the affinity of ligands with receptors of known three-dimensional structure. The ability of the model to predict the affinity of novel complexes not included in the training set was demonstrated with three independent test sets: 14 complexes of known three-dimensional structure including 3 DNA complexes, a class of compound not included in the training set, 13 HIV protease inhibitors fit to HIV-1 protease, and 11 thermolysin inhibitors fit to thermolysin.

• Scoring noncovalent protein-ligand interactions: a continuous differentiable function tuned to compute binding affinities.
Jain, A N
Journal of computer-aided molecular design, 1996, 10(5), 427-440
PMID: 8951652

Exploitation of protein structures for potential drug leads by molecular docking is critically dependent on methods for scoring putative protein-ligand interactions. An ideal function for scoring must exhibit predictive accuracy and high computational speed, and must be tolerant of variations in the relative protein-ligand molecular alignment and conformation. This paper describes the development of an empirically derived scoring function, based on the binding affinities of protein-ligand complexes coupled with their crystallographically determined structures. The function's primary terms involve hydrophobic and polar complementarity, with additional terms for entropic and solvation effects. The issue of alignment/conformation dependence was solved by constructing a continuous differentiable nonlinear function with the requirement that maxima in ligand conformation/alignment space corresponded closely to crystallographically determined structures. The expected error in the predicted affinity based on cross-validation was 1.0 log unit. The function is sufficiently fast and accurate to serve as the objective function of a molecular-docking search engine. The function is particularly well suited to the docking problem, since it has spatially narrow maxima that are broadly accessible via gradient descent.

• A fast flexible docking method using an incremental construction algorithm.
Rarey, M and Kramer, B and Lengauer, T and Klebe, G
Journal of molecular biology, 1996, 261(3), 470-489
PMID: 8780787     doi: 10.1006/jmbi.1996.0477

We present an automatic method for docking organic ligands into protein binding sites. The method can be used in the design process of specific protein ligands. It combines an appropriate model of the physico-chemical properties of the docked molecules with efficient methods for sampling the conformational space of the ligand. If the ligand is flexible, it can adopt a large variety of different conformations. Each such minimum in conformational space presents a potential candidate for the conformation of the ligand in the complexed state. Our docking method samples the conformation space of the ligand on the basis of a discrete model and uses a tree-search technique for placing the ligand incrementally into the active site. For placing the first fragment of the ligand into the protein, we use hashing techniques adapted from computer vision. The incremental construction algorithm is based on a greedy strategy combined with efficient methods for overlap detection and for the search of new interactions. We present results on 19 complexes of which the binding geometry has been crystallographically determined. All considered ligands are docked in at most three minutes on a current workstation. The experimentally observed binding mode of the ligand is reproduced with 0.5 to 1.2 A rms deviation. It is almost always found among the highest-ranking conformations computed.

## 1995

• Molecular recognition of the inhibitor AG-1343 by HIV-1 protease: conformationally flexible docking by evolutionary programming.
Gehlhaar, D K and Verkhivker, G M and Rejto, P A and Sherman, C J and Fogel, D B and Fogel, L J and Freer, S T
Chemistry & Biology, 1995, 2(5), 317-324
PMID: 9383433

BACKGROUND:An important prerequisite for computational structure-based drug design is prediction of the structures of ligand-protein complexes that have not yet been experimentally determined by X-ray crystallography or NMR. For this task, docking of rigid ligands is inadequate because it assumes knowledge of the conformation of the bound ligand. Docking of flexible ligands would be desirable, but requires one to search an enormous conformational space. We set out to develop a strategy for flexible docking by combining a simple model of ligand-protein interactions for molecular recognition with an evolutionary programming search technique.

## 1994

• Icm - a New Method for Protein Modeling and Design - Applications to Docking and Structure Prediction From the Distorted Native Conformation
ABAGYAN, R and TOTROV, M and KUZNETSOV, D
Journal of computational chemistry, 1994, 15(5), 488-506

An efficient methodology, further referred to as ICM, for versatile modeling operations and global energy optimization on arbitrarily fixed multimolecular systems is described. It is aimed at protein structure prediction, homology modeling, molecular docking, nuclear magnetic resonance (NMR) structure determination, and protein design. The method uses and further develops a previously introduced approach to model biomolecular structures in which bond lengths, bond angles, and torsion angles are considered as independent variables, any subset of them being fixed. Here we simplify and generalize the basic description of the system, introduce the variable dihedral phase angle, and allow arbitrary connections of the molecules and conventional definition of the torsion angles. Algorithms for calculation of energy derivatives with respect to internal variables in the topological tree of the system and for rapid evaluation of accessible surface are presented. Multidimensional variable restraints are proposed to represent the statistical information about the torsion angle distributions in proteins. To incorporate complex energy terms as solvation energy and electrostatics into a structure prediction procedure, a ''double-energy'' Monte Carlo minimization procedure in which these terms are omitted during the minimization stage of the random step and included for the comparison with the previous conformation in a Markov chain is proposed and justified. The ICM method is applied successfully to a molecular docking problem. The procedure finds the correct parallel arrangement of two rigid helixes from a leucine zipper domain as the lowest-energy conformation (0.5 Angstrom root mean square, rms, deviation from the native structure) starting from completely random configuration. Structures with antiparallel helixes or helixes staggered by one helix turn had energies higher by about 7 or 9 kcal/mol, respectively. Soft docking was also attempted. A docking procedure allowing side-chain flexibility also converged to the parallel configuration, starting from the helixes optimized individually. To justify an internal coordinate approach to the structure prediction as opposed to a Cartesian one, energy hypersurfaces around the native structure of the squash seeds trypsin inhibitor were studied. Torsion angle minimization from the optimal conformation randomly distorted up to the rms deviation of 2.2 Angstrom or angular rms deviation of 10 degrees restored the native conformation in most cases. In contrast, Cartesian coordinate minimization did not reach the minimum from deviations as small as 0.3 Angstrom or 2 degrees. We conclude that the most promising detailed approach to the protein-folding problem would consist of some coarse global sampling strategy combined with the local energy minimization in the torsion coordinate space. (C) 1994 by John Wiley & Sons, Inc.

## 1992

• Automated docking with grid‐based energy evaluation
Meng, EC and Shoichet, BK
Journal of computational chemistry, 1992, 13(4), 505-524

The ability to generate feasible binding orientations of a small molecule within a site of known structure is important for ligand design. We present a method that combines a rapid, geometric docking algorithm with the evaluation of molecular mechanics interaction energies.The computational costs of evaluation are minimal because we precalculate the receptor-dependent terms in the potential function at points on a three- dimensional grid. In four test cases where the components of crystallographically determined complexes are redocked, the force field' score correctly identifies the family of orientations closest to the experimental binding geometry. Scoring functions that consider only steric factors or only electrostatic factors are less successful. The force field function will play an important role in our efforts to search databases for potential lead compounds.