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