Prediction of Membrane Protein Types by an Ensemble Classifier Based on Pseudo Amino Acid Composition and Approximate Entropy

Abstract

Membrane proteins have different types. The function of a membrane protein is closely correlated with the type it belongs to. For a given membrane protein, it belongs to the following eight categories [1]: (1) type I, (2) type II, (3) type III, (4) type IV, (5) multipass, (6) lipid-chain-anchored, (7) GPI-anchored, and (8) peripheral. It is important to decide which type it belongs to because it is closely relevant to its biological function and to its interaction process with other molecules in a biological system. In this study, on the basis of the concept of pseudo amino acid (PseAA) composition originally introduced by Chou, the value of approximate entropy (ApEn) of the query membrane protein is used to integrate the complementary information. An ensemble classifier to fuse fifteen powerful individual FKNN (fuzzy K-nearest neighbor) classifiers is presented. Here each basic classifier is trained in PseAA composition of membrane protein sequences with different parameters. The success rates obtained by the ensemble classifier on both the jackknife dataset test and the independent dataset test are quite high, indicating that the ensemble classifier may become a very useful high throughput tool in identifying membrane proteins and their types.