Prediction of membrane proteins using split amino acid and ensemble classification.

摘要: Knowledge of the types membrane protein provides useful clues in deducing functions uncharacterized proteins. An automatic method for efficiently identifying proteins is thus highly desirable. In this work, we have developed a novel predicting by exploiting discrimination capability difference amino acid composition at N and C terminus through split (SAAC). We also show that ensemble classification can better exploit discriminating SAAC. study, are classified using three feature extraction several strategies. classifier Mem-EnsSAAC then best strategy. Pseudo (PseAA) composition, discrete wavelet analysis (DWT), SAAC, hybrid model employed extraction. The nearest neighbor, probabilistic neural network, support vector machine, random forest, Adaboost used as individual classifiers. predicted results learners combined genetic algorithm to form an classifier, yielding accuracy 92.4 92.2% Jackknife independent dataset test, respectively. Performance measures such MCC, sensitivity, specificity, F-measure, Q-statistics SAAC-based prediction yields significantly higher performance compared PseAA- DWT-based systems, reported so far. proposed able predict with high consequently, be very helpful drug discovery. It accessed http://111.68.99.218/membrane.