Accurate prediction of �-sheet topology is one of the most important open problems of modern computational biology. B-sheet topology prediction is an intermediate step toward the protein tertiary structure prediction. Here, a probability-based method called "BetaProbe" is presented for the �-sheet topologies prediction for proteins up to six �-strands. In this method, first, the best closed �-sheet topology for a set of �-strands is determined. Then the most frequent and stable open �­ sheet topology is found. In the proposed method, the optimum pairwise alignment between �-strands is determined using the dynamic programming approach. In finding the best closed �-sheet topology, integer optimization is used. In order to predict more accurate paired �-strands, pairwise interaction probabilities are combined with integer programming. In BetaProbe more frequent and more stable conformations have greater chances of being selected. Experimental results show that BetaProbe significantly outperforms the most recent integer programming-based method.

Keywords