Unbalance Data Processing Strategy for Protein Interaction Sites Prediction

Protein-protein interactions play essential roles in various biological progresses. Identifying protein interaction sites can facilitate researchers to understand life activities and therefore will be helpful for drug design. However, the unbalance between the negative and positive samples comes from the current definitions of interaction sites places restriction on the prediction of protein interaction sites by computational approaches. In this work, we proposed three imbalance data processing strategies to improve the performance of protein interaction sites prediction. We first proposed the extraction of relevant features based on the evolutionary conservation of amino acids to predict protein interaction sites. At the same time, three methods are proposed to deal with the imbalance of positive and negative samples in data sets. Experimental results demonstrated the effectiveness of our method.