Flexible Side Chain Models Improve Enrichment Rates in In Silico Screening

Abstract

While modern docking methods often predict accurate binding modes, affinity calculations remain challenging and enrichment rates of in silico screening methods unsatisfactory. Inadequate treatment of induced fit effects is one major shortcoming of existing in silico screening methods. Here we investigate enrichment rates of rigid-, soft- and flexible-receptor models for 12 diverse receptors using libraries containing up to 13000 molecules. For the rigid-receptor model, we observed high enrichment (EF1 > 20) only for four target proteins. A soft-receptor model showed improved docking rates at the expense of reduced enrichment rates. A flexible side-chain model with flexible dihedral angles for up to 12 amino acids (3-8 flexible side chains) increased both binding propensity and enrichment rates: EF1 values increased by approximately 35% on average with respect to rigid docking. We find on average 4 known ligands in the top 10 molecules in the rank-ordered databases for the receptors investigated.