A Unified Probabilistic Scoring Function Incorporating Multiple Sources of Information for Near-Native Discrimination of High-Resolution Protein Models

• Activity Number: 342
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Yian Ann Chen*+ and Hyun Joo and Xiaotao Qu and Ryan Day and Marina Vannucci and Jerry Tsai
• Companies: Moffitt Cancer Center and University of the Pacific and Moffitt Cancer Center and University of the Pacific and Rice University and University of the Pacific
• Keywords: template-based ; protein structure prediction ; scoring function

Abstract

We have developed a unified probabilistic scoring function incorporating the distribution of exposed polar groups, backbone dependent residue volumes, and backbone dependent chi1 angles for template-based protein structure prediction. Using an extensive set of molecular dynamics simulations, we estimated the probabilities of the generated candidate structures for 41 CASP7 targets. On average, 253 candidate structures were generated for each target. For each target set, we classified the candidates into high- or low- score groups when ranked in the top 5% or the remaining 95%. There is a significant difference in mean RMSD between the high- and low-score groups (paired-t = 4.6, p < 10^-4) with the mean difference of 1.02 Å (+-0.22 Å). Our score provides a promising way to select near native structures in a unified framework within the regime of template based structure prediction