Predicting Protein-Protein Interactions from Protein Sequence Information Using Dual-Tree Complex Wavelet Transform.

Protein-protein interactions (PPIs) play major roles in most biological processes. Although a number of high-throughput technologies have been established for generating PPIs, it still has unavoidable problems such as time-consuming and labor intensive. In this paper, we develop a novel computational method for predicting PPIs by combining dual-tree complex wavelet transform (DTCWT) on substitution matrix representation (SMR) and weighted sparse representation-based classifier (WSRC). When predicting PPIs of Yeast and Human datasets, the proposed method obtained remarkable results with average accuracies as high as 97.12% and 97.56%, respectively. The performance of the proposed method is obviously better than the existing methods. Furthermore, we compare it with the superior support vector machine (SVM) classifier for further evaluating the prediction performance of our method. The promising results illustrate that our method is robust and stable for predicting PPIs, and it is anticipated that it would be a useful tool to predict PPIs in a large-scale.