Prediction of membrane protein types by exploring local discriminative information from evolutionary profiles

Membrane protein is a pivotal constituent of a cell that exerts a crucial influence on diverse biological processes. The accurate identification of membrane protein types is deeply essential for revealing molecular mechanisms and drug development. Primarily, several traditional methods were exploited to classify these types. However, experimental methods are laborious, time-consuming, and costly due to rapid exploration of uncharacterized protein sequences generated in the postgenomic era. Hence, machine learning-based methods are more indispensable for reliable and fast identification of membrane protein types. A variety of state-of-the-art investigations have been elucidated to improve prediction performance, but predictive validity is still insufficient. Motivated by this, we designed a promising sequential support vector machine based predictor called TargetHMP to predict types of membrane proteins. We captured the local informative features by exploring evolutionary profiles through a novel method called the segmentation-based pseudo position-specific scoring matrix (Seg-PsePSSM). TargetHMP attained high accuracy of 94.99%, 93.48%, and 90.36% on the S₁, S₂, and S₃ datasets, respectively, using a vigorous leave-one-out-cross-validation test. The results indicate that the performance of the proposed method outperformed prior predictors. We expect that the proposed approach will help research academia in general and pharmaceutical drug discovery in particular.