β-sheet Topology Prediction Using Probability-based Integer Programming

Abstract

β-sheet topology prediction is a major unresolved problem in modern computational biology. It is a challenging intermediate step toward the protein tertiary structure prediction. Different methods have been provided to deal with the problem of determining the β-sheet topology. Here, ab-initio probability-based methods called "BetaProbe1" and "BetaProbe2" are utilized to specify the β-sheet topology. In these methods, the stability and the frequency of β-strand pairwise interaction and β-sheet conformation are spotted. To predict more frequent interactions between β-strand pairs, besides pairwise alignment probability, the probability of occurring β-strand pairwise interaction is considered to compute the score of the interactions. Furthermore, to determine the β-strand pairwise alignment probability more accurately, a dynamic programming approach is utilized. In addition, the integer programming optimization is combined with the probabilities of β-strand pairwise interactions to determine the β-sheet topology. Moreover, the β-sheet conformation probability is considered to give better chances to more observed conformations for selection. Experimental results show that BetaProbe1 and BetaProbe2 significantly outperform the most recent integer programming-based method with respect to β-sheet topology prediction.

Keywords

• β-sheet topology prediction
• integer programming
• dynamic programming
• pairwise alignment