Exploring and Exploiting the Biomolecular Structure and Function with Machine Learning: Biodiversity and Beyond

We can think about biodiversity from an evolutionary and organismal perspective, cataloging and modeling biodiversity of interacting species. Another useful way to think about biodiversity is to think of the diversity of biological sequences and structures. Multiple large-scale public and private efforts have sequenced, using automated DNA sequencing technology, a vast number of genomes and fragments of genomes that span the tree of life. Within those genomes are sequences that code for a very large assortment of molecular structures (RNA and protein structures) and functions (everything from catalysts to proteins that process cellular information).

In this Presidential Lecture, Richard Bonneau will discuss new machine-learning methods for characterizing these biomolecules. He will discuss how these methods lead to methods for designing new molecules not seen before in nature. The methods presented will lie at the intersection of modeling the physics of protein structure and the evolution of protein sequences, using the power of machine learning to integrate these two disparate but equally important ways of capturing biological structure-function relationships. A key objective of this work is to leverage models that, having trained on bio-diverse sequences and structures, can be used to design new therapies. Lastly, Bonneau will discuss recent machine learning-powered drug discovery work that aims to dramatically improve the effectiveness, safety, and cost of new therapies.

Bonneau leads Prescient Design, a molecular design accelerator at Genentech (a member of the Roche Group) that pioneers new methods for combining machine learning and molecular modeling. Bonneau’s research spans multiple levels of biological structure learning and modeling biological networks to predicting and designing macromolecular and biomimetic structure. Rich received his Ph.D. at the University of Washington, Seattle, where he pioneered new methods to predict biomolecular structure as a member of the Rosetta project. At Genentech Research and Early Development (gRED), Bonneau and his team build new methods for applying machine learning to design, molecular composition, function and interfaces in ways that span all drug modalities.