Bayesian Nonparametric Methods for Protein Structure Prediction

• Activity Number: 401
• Type: Topic Contributed
• Date/Time: Tuesday, August 2, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Kristin Patricia Lennox*+ and David B. Dahl and Marina Vannucci and Ryan Day and Jerry W. Tsai
• Companies: Lawrence Livermore National Laboratory and Texas A & M University and Rice University and University of the Pacific and University of the Pacific
• Keywords: Dirichlet process mixture model ; Bayesian nonparametrics ; protein structure prediction ; density estimation

Abstract

The protein structure prediction problem consists of determining a protein's three-dimensional structure from the underlying sequence of amino acids. For template-based modeling, a target is assumed to be structurally similar to other proteins with known structure. We present statistical methods for incorporating information about template protein structures into searches of protein conformation space. The general strategy is to identify a simplified representation of protein structure and then develop a statistical model for nonparametric density estimation. This process is used to first model backbone torsion angles at individual protein sequence positions, then extended to simultaneous modeling at multiple positions. The final modification is to incorporate information about protein side chain positions into the existing backbone model. Analysis of the protein structure data affords the opportunity to explore various extensions to the standard Bayesian density estimation framework, including the incorporation of priors into otherwise nonparametric models and a method for modeling dependence which is an alternative to nonparametric copulas.