Subcellular Locations Prediction of Proteins Based on Chaos Game Representation

To understand the functions of various proteins, it would be helpful to obtain information about their subcellular locations. With the rapid accumulation of newly found protein sequence data in databanks, it would be worthwhile to develop a fast computational prediction method to identify protein's subcellular location. In this paper, we considered 4 subcellular locations of proteins from rice: chloroplast, cytoplasmic, integral membrane protein and nucleus. Our data set is the proteins with known locations from the SWISS-PROT and TrEMBL database. We introduced the Chaos Game Representation (CGR) of protein to transform the protein sequence into the numerical vector, instead of the quasi-amino acid composition. Furthermore, we added two dimensions in the end based on the amino acid's physics chemistry properties. The results show that the Chaos Game Representation is better than the amino acid composition, and the new characters can improve the accuracy obviously.