Title:Â Transforming Early-Stage Drug Discovery Through integration of Machine Learning and Structure-Based Approaches
Abstract:Â Recent advances in computational approaches have transformed early-stage drug discovery, enabling more efficient identification of promising therapeutic candidates. This talk explores the synergistic application of machine learning and physics-based methods for hit discovery and lead optimization, with a focus on targeting complex protein-protein interfaces. We present our computational approach through multiple case studies, highlighting a key innovation: the development of pharmacophore models through systematic peptide mutation to non-natural amino acids. This strategy is demonstrated in targeting the challenging E6-P53 protein-protein interaction, where mapping critical binding interactions through peptide modifications led to a comprehensive pharmacophore model that guided small molecule PPI inhibitor discovery. Our platform integrates active learning strategies to efficiently sample the expanded chemical landscape. Model predictions inform experimental prioritization, while new data continuously refines our computational models. Through these case studies, we examine both the transformative potential and current limitations of computational methods in modern drug discovery.
Hosted by Pratyush Tiwary
Visit go.umd.edu/BIPH-seminar-subscribe to be added to the email list.