This is an online seminar. Registration is required.
【Molecular Informatics Team】
【Date】2024/September/5(Thu) 14:30-15:30(JST)
*【Speaker】Dr. Andrejs Tucs, University of Tokyo *
Title: Generative modeling for drug design applications
Abstract:
Conventional drug design requires human expertise to assist candidate molecule proposal, synthesis, and evaluation. This process can be cost/time demanding, limited in amount of molecules that can be experimentally probed. Generative modeling provides an alternative to this by reformulating drug design as an inverse design problem. In the first part of the talk I will discuss Generative Adversarial Network (GAN) and Variational AutoEncoder (VAE) use in this context. As I will show incorporation of various biological priors can substantially expand both model capabilities. Their performance can be further enchanced by incorporating emerging quantum computing technologies such as quantum annealing. In the second part I will discuss low computational complexity constrained Monte-Carlo geometric simulation techique for molecule conformational sampling with use cases and how it can implicitly/explicitly assist drug design generative modeling pipelines.
