A Novel Statistical Potential for Protein Beta-Sheets Prediction
Master of Science
One of the most long-term challenging problems in biophysics studies for both computational scientists and experimentalists is protein structure prediction, whose goal is to obtain three-dimensional native protein structure from one-dimensional sequence. In protein structure prediction problems, a fundamental problem is Beta-sheets structure prediction. Though more than 85% of experimentally solved proteins contain Beta-sheet structures, limited methods have been found to rapidly and accurately predict the folded conformations. In this study, we proposed a novel statistical potential, named NP-Beta, to predict the protein Beta-sheet structure only based on the sequence information. We included three kinds of potential terms in NP-Beta, i.e. the self-packing term, the pair interacting term and the lattice term. The number of hydrogen bonds in Beta-sheets is also considered as a potential component, corresponding to a global penalty of the potential function. Computational tests show that the new statistical potential has an outstanding performance on native structure recognition from decoys comparing to the Beta-sheet specific potentials in literature. We will apply the potential to improve the prediction of Beta-strand arrangement and registration for beta proteins.
Statistical potential; Beta-Sheets