MemHyb: predicting membrane protein types by hybridizing SAAC and PSSM.

About 50% of available drugs are targeted against membrane proteins. Knowledge of membrane protein's structure and function has great importance in biological and pharmacological research. Therefore, an automated method is exceedingly advantageous, which can help in identifying the new membrane protein types based on their primary sequence. In this paper, we tackle the interesting problem of classifying membrane protein types using their sequence information. We consider both evolutionary and physicochemical features and provide them to our classification system based on support vector machine (SVM) with error correction code. We employ a powerful sequence encoding scheme by fusing position specific scoring matrix and split amino acid composition to effectively discriminate membrane protein types. Linear, polynomial, and RBF based-SVM with Bose, Chaudhuri, Hocquenghem coding are trained and tested. The highest success rate of 91.1% and 93.4% on two datasets is obtained by RBF-SVM using leave-one-out cross-validation. Thus, our proposed approach is an effective tool for the discrimination of membrane protein types and might be helpful to researchers/academicians working in the field of Drug Discovery, Cell Biology, and Bioinformatics. The web server for the proposed MemHyb-SVM is accessible at http://111.68.99.218/MemHyb-SVM.