Regina Barzilay, MIT
Zoom link: https://berkeley.zoom.us/j/94710860875
Abstract:
The first generation of AI chemistry models was based on 2D models of molecular structure. Given sufficient amounts of training data, these models can be effectively used to predict a range of biochemical properties, to perform lead optimization and to support full de-novo design. However, the utility of these models deteriorates in many realistic scenarios due to data sparsity and a distributional shift between the training and testing data. In this talk, I will discuss how infusing these models with biological information improves their ability to generalize. This will range from incorporating mechanism of action information to geometric modeling of protein/ligand and protein/protein interactions.
Bio: Regina Barzilay is a School of Engineering Distinguished Professor for AI and Health in the Department of Electrical Engineering and Computer Science and a member of the Computer Science and Artificial Intelligence Laboratory at the Massachusetts Institute of Technology. She is an AI faculty lead for Jameel Clinic, an MIT center for Machine Learning in Health. Her research interests are in machine learning models for molecular modeling with applications to drug discovery and clinical AI. She also works in natural language processing. She is a recipient of various awards including the NSF Career Award, the MIT Technology Review TR-35 Award, Microsoft Faculty Fellowship and several Best Paper Awards at NAACL and ACL. In 2017, she received a MacArthur fellowship, an ACL fellowship and an AAAI fellowship. In 2020, she was awarded the Squirrel AI Award for Artificial Intelligence for the Benefit of Humanity. She received her PhD in Computer Science from Columbia University, and spent a year as a postdoc at Cornell University. Prof. Barzilay received her undergraduate degree from Ben-Gurion University of the Negev, Israel.