UCB Protein Structure Prediction Project

People: Na Yang, Jingwei Guo, Tianqi Wang and Wendi Heinzelman

Sponsors:


Figure 1: Protein 3D structure plotted in MATLAB with atoms on one residue labeled
Project overview: Protein structure prediction has been widely used in drug efficacy prediction, and drug side-effect prediction. The goal of this project is to discover connections between bioinformatics and wireless communication, thus tackling the protein structure prediction problem in an interdisciplinary way.

Given the protein backbone structure, we aim to find side-chain conformations that minimize the protein structure's overall energy. Figure 1 shows a protein 3D structure with one side-chain (residue) labeled. Discrete side-chain conformations are known as rotamers. We predict side-chain conformation from a statistic perspective by using previously solved protein structures. Backbone-dependent rotamer libraries present rotamer conformations on the local backbone conformation as defined by the backbone dihedral angles Φ and Ψ. We have studied the 2010 backbone-dependent rotamer library by Dunbrack (http://dunbrack.fccc.edu/bbdep2010/) that uses an adaptive kernel on the backbone-dependent rotamer angle probability distributions. We have generated the smoothed rotamer probabilities based on the existing library, essentially replicating the work in Dunbrack’s paper. Figure 2 shows a smoothed rotamer probability of Χ 1 = g + for the residue type Serine. Using Dunbrack’s scoring function, which models energy components in the protein, we are developing a new sampling strategy, with the goal being to provide the same results with much less computation.
Figure 2: Smoothed rotamer probability of for the residue type Serine.