2Gram Features Based Prediction of Membrane Protein Types Using Ensemble Classifiers Methods
18 Pages Posted: 25 Nov 2020 Last revised: 4 Dec 2020
Date Written: November 19, 2020
Membrane proteins are an essential type of protein used in a cell as receptors, channels, and energy transducers. Moreover, they do many of the functions imperative to the survival of the cell. Therefore, it is necessary to build up machine learning methods to correctly identify membrane protein types to recognize protein function, disease occurrence, and drug therapy design. So the function prediction of membrane protein performs a key task. In this paper, membrane protein types are detected by using 2-gram and sequence length feature. The predictor ensemble classifiers used to predict a membrane protein, with sequence length encodes single descriptor value and 2-gram exchange group, encodes multilabel feature types. Were, 2-gram special kinds of feature type focus on its different modes of protein information to emulate the fundamental functions that are enormously unknown in complex protein sequences and find a relation between residues. The proposed method avoids biasing among the differentiation between types of membrane proteins. Ensemble classifiers well handle the imbalanced protein datasets.
Keywords: Membrane Protein; SVM; AdaBoost; RUSBoost; Random Forest; Extratree; 2Gram Exchange Group.
Suggested Citation: Suggested Citation