IT seminar: Geometric deep learning for structural bioinformatics research

Speaker: Vincent Mallet

Abstract:

Protein structure is key to understand biological processes such as binding to other compounds, the foundation of drug discovery. Determination of the structure of a protein from images and prediction of the binding is therefore a problem of paramount importance. Deep learning methods can leverage vast amounts of data and have proven efficient on many problems already. I want to extend these methods to geometric data, such as surfaces and cloud of points. I want then to use the result to solve the structure determination and binding prediction problems.