Drug-enzyme interaction studies using computational models

Abstract

How accurately can computational models predict drug-enzyme binding interactions? This research evaluated molecular docking predictive performance for estimating drug-enzyme binding affinities. A systematic comparison of four docking software platforms was conducted at Mayon Institute between May 2016 and December 2017. The evaluation dataset comprised 45 drug-enzyme complexes with experimentally determined binding affinities spanning kinase, protease, and phosphatase inhibitors. Predicted binding energies demonstrated strong correlation with experimental values (R² = 0.87, RMSE = 1.24 kcal/mol). Glide demonstrated highest accuracy with 88% correct pose prediction, while AutoDock Vina achieved fastest computation at 92% relative speed. Predicted inhibition constants showed agreement within one order of magnitude for 89% of compounds. These findings establish benchmarks for computational prediction of drug-enzyme interactions.