Predicting Protein-Protein Interactions from Protein Sequence Using Locality Preserving Projections and Rotation Forest.

Protein-protein interactions (PPIs) play an important role in nearly every aspect of the cell function in biological system. A number of high-throughput technologies have been proposed to detect the PPIs in past decades. However, they have some drawbacks such as time-consuming and high cost, and at the same time, a high rate of false positive is also unavoidable. Hence, developing an efficient computational method for predicting PPIs is very necessary and urgent. In this paper, we propose a novel computational method for predicting PPIs from protein sequence using Locality Preserving Projections (LPP) and Rotation Forest (RF) model. Specifically, the protein sequence is firstly transformed into Position Specific Scoring Matrix (PSSM) generated by multiple sequences alignments. Then, the LPP descriptor is applied to extract protein evolutionary information from. Finally, the RF classifier is adopted to predict whether the given protein pair is interacting or not. When the proposed method performed on Yeast and H. pylori PPIs datasets, it achieved the results with an average accuracy of 92.52% and 91.46%, respectively. To further verify the performance of the proposed method, we compare the proposed method with the state-of-the-art support vector machine (SVM) and get good results. The promising results indicated the proposed method is stable and robust for predicting PPIs.