Authors: Fransisco Serna
Primary Advisor: Claudio N. Cavasotto, PhD (co-author)
Committee Members:
Masters thesis, The University of Texas School of Biomedical Informatics at Houston.
Abstract:
G Protein-Coupled Receptors (GPCRs) comprise the largest family of transmembrane receptors, with much research devoted to solving protein structures. In this contribution, we explore structure- and ligand-based virtual screening methods for drug discovery. We use a machine-learning algorithm, a Laplacian-modified naïve bayes classifier, to predict and rerank a list of compounds from high-throughput docking (HTD) scoring. We then evaluate the overall effectiveness to enrichment and the effectiveness of the Laplacian-modification. Our results demonstrate superior enrichment gains using the Laplacian-modified naïve bayes classifier for both structure- and ligand-based methods. These methods show that with no, or little a priori knowledge of binding activity, enrichment can be significantly increased given meaningful information is supplied to the classifier.