Physics Dept. Seminar
November 29, Monday
Physics-Based Machine Learning of Protein Structure and Function
Prof. Guillaume Lamoureux
Dept of Chemistry, Rutgers University- Camden
(Biophysics, Host: Dias)
Room: ECE 202
**TIME: 11:40am – 12:40pm, door opening at 11:30am
Proteins are the main building blocks of living organisms. To understand any biological process in detail, one needs to know the structure, function, and dynamics of all proteins involved. Despite considerable advances in proteomics, structural and functional data are available only for a small fraction of all known proteins — and an even smaller fraction of all *possible* proteins (those that were not observed during evolution but that are accessible through protein engineering). To alleviate this severe data imbalance, computational methods based on sound physical principles are needed.
In this talk, I will present an overview of the current state of machine learning of proteins and will discuss our recent contributions to the development of unified “sequence-to-structure-to-function” models based on deep neural networks. Following an end-to-end learning approach, these models aim to predict how proteins assemble and interact with one another by discovering the data representations most useful for the tasks of predicting function from structure. I will also discuss some of the new research avenues opened by deep learning approaches to biomolecular data.