NYAGIM Event: Dr. Pratyush Tiwary
Details
Can AI help learn and predict biomolecular conformations and dynamics?
Molecular dynamics has been useful in understanding generic biomolecular receptors and designing drugs to bind them. However robust methods for correctly sampling slow biomolecular dynamics remain elusive. In this talk I will show that certain flavors of AI integrated with MD can help us make a big dent in this problem. This requires close integration of AI with old and new ideas in statistical mechanics. I will talk about some such methods developed by my group [1] using different flavors of AI including information bottleneck, denoising probabilistic models and recurrent neural networks. I will demonstrate the methods on different problems, where we predict mechanisms at timescales much longer than milliseconds while keeping all-atom/femtosecond resolution. These include ligand dissociation from flexible kinase/riboswitch and ligand permeation through membranes. Time/preprint permitting I might also talk about some exciting new work on binding site prediction and recovering Boltzmann diversity from Alphafold2.
