Evaluation of Docking Calculations on X-Ray Structures Using CONSENSUS-DOCK

Abstract

NoteDuring the past decade, in silico approaches for seeking active compounds to target proteins have become more popular in drug discovery due to significant advances in computer hardware and software. There are a large number of docking programs (e.g., DOCK, 1) AutoDock, 2) GOLD, 3) GLIDE, 4) Ph-DOCK 5) ), and many applications of these programs have been reported by various research groups. In major pharmaceutical companies, in silico approaches have been used to solve complicated drug design problems, leading to an increased appreciation and understanding of in silico approaches among medicinal chemists, pharmacological scientists and other researchers. 6,7)We have applied several in silico approaches to advance our drug discovery efforts, including structure-based virtual screening (SBVS), ligand-based drug design (LBDD), fragment-based drug design (FBDD), and assessment of proteinprotein interactions (PPI). 8) We have been particularly successful in efficiently finding active compounds for various target proteins (e.g., GPCR, kinase) by SBVS, and have generally used an in-house docking program, CONSENSUS-DOCK.9) CONSENSUS-DOCK is a customized version of the DOCK4 program in which three scoring functions (DOCK4, FlexX 10) and PMF 11) ) and consensus scoring have been implemented. We previously reported the advantages of CONSENSUS-DOCK over DOCK4 in SBVS for a target protein, death-associated protein kinases (DAPKs).9) To confirm the advantages of CONSENSUS-DOCK by using other types of proteins, we compare the docking calculation results obtained using CONSENSUS-DOCK and DOCK4 for 16 Xray crystal structures in this paper. We report that CONSEN-SUS-DOCK provided better docking calculations than DOCK4 for most of the 16 X-ray structures chosen from the Protein Data Bank (PDB). 12)In CONSENSUS-DOCK, three scoring functions (DOCK4, FlexX and PMF) are implemented in the DOCK4 program and consensus scoring 13) is performed in the pose selection, and the compounds are ranked to reduce the weaknesses of each scoring function. DOCK4 is a force field based docking program developed by Kuntz et al. 1) and is used for automated molecular docking of flexible molecules. DOCK4 uses electrostatic and van der Waals interactions evaluated over a grid to calculate the binding energy of a docked conformation. FlexX is an empirical scoring function and is implemented in the SYBYL package. Standard parameters of FlexX's scoring function are used for iterative growing and subsequent scoring of docking poses. PMF is a knowledge-based scoring function that combines the accuracy of empirical scoring functions with the advantage of higher generality and therefore wider applicability.Various kinds of consensus scores have been used by sev- We are participating in the challenge of identifying active compounds for target proteins using structurebased virtual screening (SBVS). We use an in-house customized docking program, CONSENSUS-DOCK, which is a customized version of the DOCK4 program in which three scoring functions (DOCK4, FlexX a...

Are predefined decoy sets of ligand poses able to quantify scoring function accuracy?

“…If no information on the scoring function performance for a particular target is available, a worthwhile option can be to look at the scoring function performance on highly similar targets provided that any exist. In case no related targets exist, the use of consensus scoring schemes may be the protocol of choice due to their relatively high robustness to outliers [23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38]. Given the lucky situation that structural and/or affinity data for the target at hand is available, assessing the performance of these scoring functions within the same docking package is obviously the most effective way for making an informed decision about which one to choose in the discovery process.…”

“…As shown by O'Boyle et al [22], the performance of a specific scoring function is almost always better if it is used to rescore poses generated using another scoring function than if the function is directly used to generate the poses. Many studies have shown that consensus docking schemes benefit from this observation [23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38]. In order to evaluate the pose prediction performance of the scoring functions more objectively, the conformational space should be sampled as accurately as possible.…”

Molecular Docking and Structure-Based Drug Design Strategies

“…This can be very useful, as it combines the advantages and simultaneously attenuates the shortcomings of each method [64]. Examples of consensus scoring functions are MultiScore, X-Cscore, GFscore, SCS, SeleX-CS and CONSENSUS-DOCK [65][66][67][68][69][70]. Table 2 provides a list of several scoring functions implemented in the most frequently used molecular docking programs.…”

Latest developments in molecular docking: 2010–2011 in review

“…Pose prediction: RMSD [94] Glide XP, Glide SP, Surflex, FlexX, GOLD 4 targets: angiotensin-converting enzyme (ACE), cyclooxygenase 2 (COX-2), thrombin and human immunodeficiency virus 1 (HIV-1) protease; 50 actives and 950 decoys from the MDL Drug Data Report (MDDR) for each target [204] GOLD: GoldScore, ChemScore, ASP; AutoDock; Surflex-Dock; FRED: Shapegauss, PLP, CGO, Chemgauss 2, Chemgauss 3 4 sarco/endoplasmic reticulum calcium ATPase (SERCA) complexes for pose prediction; 22 SERCA inhibitors for affinity prediction Pose prediction: RMSD; affinity prediction: score vs. logIC 50 [205] PDBbind (610)…”

“…Other examples of consensus scoring include VS against kinases; [29] VoteDock development; [17] a combination of quantitative structure-binding affinity relationship (knowledge based) and MedusaScore (force field based); [91] a combination of MCSS and GOLD docking with MM/GBSA rescoring for fragments; [32] a combination of AutoDock4 and AutoDock Vina; [92] a combination of DOCK4, [93] FlexX, and PMF; [94] and HarmonyDOCK. [95] Machine-learning approaches to scoring function development One of the postulated weaknesses of scoring, resulting in poor affinity prediction, is the assignment of a common (i.e., system-independent) set of weights to the individual functional terms and the incorrect assumption that these weights are additive in their contribution to binding affinity.…”