ABSTRACT 

This talk explores how the integration of biomedical knowledge can enhance generative artificial intelligence for drug discovery. Recent generative models have demonstrated strong capabilities in molecular design, yet they operate largely disconnected from the extensive scientific knowledge systematically encoded in databases and literature. We develop frameworks that embed structured representations of chemical relationships, biological pathways, and therapeutic contexts into generative models, enabling the synthesis of drug candidates that are not only chemically valid and synthesizable but also aligned with specific therapeutic targets and biological mechanisms. This approach moves beyond heuristic-driven methods by grounding generation in explicit biomedical understanding. When augmented with embeddings from large-scale knowledge graphs, these models produce molecules with improved binding affinities, predicted efficacy, and relevance to disease mechanisms while maintaining flexibility to address both single-target and multi-target therapeutic scenarios. The results demonstrate that connecting generative processes to accumulated scientific knowledge improves the design of therapeutically relevant compounds. By bridging the gap between pattern-based generation and domain expertise, this work establishes knowledge-enhanced frameworks as practical tools for rational drug design and therapeutic development.

BIOGRAPHY

Aditya Malusare is a Ph.D. candidate at the Edwardson School of Industrial Engineering at Purdue University. He earned his M.S.in Physics from Purdue University in 2024, and his B.Tech. in Engineering Physics from the Indian Institute of Technology, Bombay.