Predicting subcellular localization of proteins with multiple sites using threshold ml-KNN

Abstract

Predicting the appropriate protein subcellular localization has attracted much attention in the field of bioinformatics for determining the cellular function of proteins. Several traditional biochemical experimental methods have been developed to determine the protein subcellular localization is expensive and time-consuming and during the last decade, many computational based methods have been developed to predict the protein subcellular localization in different organisms. However, most of the methods succeeded to predict proteins in only one subcellular location but there are many proteins which have two or more subcellular locations. To predict subcellular localization of protein with multiple sites we used Threshold-MLKNN (multi-label k-Nearest Neighbours algorithm). Threshold-MLKNN performs much better than MLKNN and produces better prediction accuracy in much lesser time.