Wednesday, December 2, 2009
3:00 - 4:00
Room 457 Blocker
Texas A&M University
Department of Statistics
Bayesian Nonparametric Methods for Protein Structure Prediction
The protein structure prediction problem consists of determining a protein's three-dimensional structure from the underlying sequence of amino acids. One approach is template-based structure prediction. In the template-based framework, a new protein is suspected to be structurally similar to other proteins with known structure. There are many potential applications for statistics in this area, and this presentation will focus on strategies for generating candidate structures for a target protein based on the distribution of closely related protein structures.
This talk addresses two critical issues for generating candidate structures: protein backbone prediction and the placement of amino acid side-chains relative to the backbone. Both problems can be addressed using Bayesian nonparametric density estimation techniques. There have been a number of advances within the Bayesian paradigm prompted by unique problems inherent in protein structure prediction including methods for dealing with angular data, sparse data, and modeling association for disparate data types. These are presented in the context of the overarching structure prediction problem.