Machine learning approaches for customized docking scores: Modeling of inhibition of Mycobacterium tuberculosis enoyl acyl carrier protein reductase

Abstract

Machine learning algorithms were used for feature selection and model generation of customized docking score functions for known inhibitors of Mycobacterium tuberculosis enoyl acyl carrier protein reductase. The features included small molecule descriptors derived from MOE, Accord, and Molegro as well as in silico docking energies/scores from GOLD and Autodock. The resulting models can be used to identify key descriptors for enoyl acyl carrier protein reductase inhibition and are useful for high-throughput screening of novel drug compounds. This paper also evaluates and contrasts several strategies for model generation for quantitative structure-activity relationships.