MPI-NAT Seminar Series: Molecular Science in the Age of AI

Deep learning, aka AI, has already made profound impact in the sciences, as highlighted by the 2024 Nobel prizes in Chemistry and Physics. In this talk I will discuss how deep learning can continue to disrupt some of the fundamental challenges in the molecular sciences, e.g., by solving the Schrödinger equation with unprecedented accuracy or emulating large molecular Systems at unprecedendet throughput. I will focus on discovering biomolecular mechanisms that lead to function at scale by a biomolecular emulator (BioEmu). BioEmu can predict the flexibility and conformational states of proteins, emulate the distributions generated by long-term molecular dynamics (MD) simulations orders of magnitude faster and predict protein stabilities consistent with experiment. I will highlight opportunities for drug discovery and discuss limitations of faithful physics representation and training data availability and how they might be overcome.