Abstract:  Structure-based drug design has transformed modern therapeutic strategies by leveraging computational techniques, enabling the screening of millions of compounds within feasible timeframes. However, the hit rate of this approach remains limited, primarily due to the high number of false positives. While molecular dynamics (MD) simulations are highly accurate to study ligand binding affinities, their computational cost makes it impractical to study more than a few hundred compounds. By integrating molecular docking, MD simulations, and deep learning techniques, we aim to develop a platform capable of efficiently screening large compound libraries with sufficient predictive power.

In this talk, I will first illustrate how my past computational work has contributed to the early-stage design of lead molecules, with a specific focus on developing inhibitors against the Human Immunodeficiency Virus-1 (HIV-1) accessory protein Nef. I will highlight how traditional molecular docking methods can be refined to generate novel inhibitors with high synthetic accessibility. Additionally, I will introduce enhanced sampling techniques for MD simulations that improve our understanding of molecular specificity and binding mechanisms.

Finally, I will discuss how we incorporate deep learning methods into our ongoing research on the human cardiac myosin protein and their potential application in treating hypertrophic cardiomyopathy. These advancements illustrate how the synergy between structural biology, physics-based molecular simulations, and artificial intelligence is reshaping drug discovery.