Molecular docking- and genetic algorithm-based approaches to produce robust 3D-QSAR models

Abstract

Conformational alignment is a crucial step in 3D-Quantitative structure–activity relationship (QSAR) modeling, e.g., Comparative Molecular Field Alignment (CoMFA). Different alignment methods including alignment on an appropriate template conformation, field fit alignment, database alignment, pharmacophore-based alignment, atom fit, and docking-based alignment are preferred choices for 3D-QSAR modeling. Our experience suggested that high quality 3D-QSARs are not guaranteed if docking/scoring or combination of both used as a pre-alignment criteria. Therefore, there is a need to investigate whether the widely used docking methods and scoring functions are the best choices for envisaging the alignment hypothesis. In the current study, a comparative evaluation of docking algorithms and scoring functions was performed and evaluated. For the purpose of evaluation a subset of widely used docking algorithms were used for docking based alignment in CoMFA. It was concluded that no docking or scoring method was able to produce the robust QSAR model (reflected by low q2 values) at least for the case of the dataset used in current study. In order to improve the relatively poor q2 values, we performed an evolutionary method and the q2 value was optimized using genetic algorithm based conformational selection approach. A dataset of Glycogen Phosphorylaseb (E.C.2.4.1.1) inhibitors was used to apply the protocol.