Protein β-sheet structure prediction is one of the most critical and challenging intermediate steps towards the full identification its tertiary structure due to the presence of non-local interactions. As a result, defining scoring functions for different levels of β-sheet structure prediction (i.e., contact map, pairing β-strands and their arrangements in β-sheets) is a critical task. Thus, the optimum conformation, which is determined by the existing functions, have some inaccuracies. Here, to achieve more reliable solution, we propose an algorithm that utilizes a structural consensus of superior conformations. The algorithm is proposed in such a way that each selected conformation contributes in making the result based on its rank. For this purpose, we create a complete directed weighted strand pairing graph and then find the maximum spanning subgraph of the presented graph as an output of the problem. Experimental results showed significant improvements in the prediction accuracy.

Keywords