Topics in Generative Modeling Methods in Protein Modeling and Design

Catalog description

The past five years have brought breakthroughs in computational protein modeling and design as a result of new machine learning methods, especially generative models. This course surveys recent papers on the development and application of methods in this space. The focus is on (1) applications: protein structure prediction (AlphaFold), sequence and structure generation, similarity search, and ensemble modeling, and (2) enabling methods: graph neural networks, transformers, autoencoders, and diffusion models. The course centers on discussion and class participation. An open-ended final project provides an opportunity to relate course topics to thesis interests or to replicate or apply the results of a recent paper. Enrollment is limited to PhD students from any department or requires permission of the instructor.