RP-FIRF: Prediction of Self-interacting Proteins Using Random Projection Classifier Combining with Finite Impulse Response Filter

Abstract

The self-interacting proteins (SIPs) plays a significant part in the organism and the regulation of cellular functions. Thence, we developed an effective algorithm to predict SIPs, named RP-FIRF, which merges the Random Projection (RP) classifier and Finite Impulse Response Filter (FIRF) together. More specifically, the Position Specific Scoring Matrix (PSSM) was firstly converted from protein sequence by exploiting Position Specific Iterated BLAST (PSI-BLAST). Then, we obtained the same size of matrix by implementing a valid matrix multiplication on PSSM, and applied FIRF approach to calculate the eigenvalues of each protein. The Principal Component Analysis (PCA) approach is used to extract the most relevant information. Finally, the performance of the proposed method is performed on human dataset. The results show that our model can achieve high average accuracies of 97.89% on human dataset using the 5-fold cross-validation, which demonstrate that our method is a useful tool for identifying SIPs.