3D-Equivariant Graph Neural Networks for Refining and Evaluating Protein Structures
October 21 @ 11:00 – 12:00 CEST
Speaker:
Jianlin Cheng, Department of Electrical Engineering and Computer Science at the University of Missouri, Columbia, USA.
Title
3D-Equivariant Graph Neural Networks for Refining and Evaluating Protein Structures
Abstract
Deep learning is revolutionizing the prediction of protein structure and is close to solve this grand challenge hanging over the scientific world for many years. In this talk, I will describe how this technology emerged in the field, how it overcame various technical hurdles to reach a high accuracy of predicting protein structures as demonstrated by AlphaFold2, and where it is going now. I will present our latest work of applying 3D-equivariant graph neural networks with self- attention to evaluate and refine protein structural models. Our experiments demonstrate that 3D-equivariant graph network networks that are robust against the rotation and translation of 3D objects can evaluate and improve the quality of protein structures more effectively than the existing methods.
Google Calendar 
