Using Different Feature Combination Methods to Study Multisite Protein Sub-cellular Localization Prediction

Abstract

Multisite protein sub-cellular localization prediction has become the hot topic relating biological information in recent years. Quite a lot of researchers have researched multisite protein sub-cellular localization for a long time. However, the accuracy still needs to be improved. As one of the researchers, I should explore new methods to improve the prediction accuracy. I choose Gpos-mPLOC data set in this paper. In addition, combining the pseudo amino acid composition, position vector and entropy density three effective feature extraction methods arbitrarily to extract protein features. Then, putting these features into multi-label k nearest neighbor classifier to predict protein sub-cellular location. The experiment proves that different feature combination methods can result in different prediction accuracy through the Jack-knife test and I can choose the best feature combination method to predict multisite protein sub-cellular location.