Identifying Species based on Relative Synonymous Codon Pair Usage combining K - means and SVM

Abstract

Species identification is beneficial for many aspects of life and scientific research, but the experiment method based on biochemistry may be subjective and inaccuracy in several cases. In order to solve this problem, searching genes in the database is one of the most effective and accurate methods for identification of the Bacillus. However, in the case of the incomplete database, the searching algorithm cannot identify genes which are not in the database. Thus, in this research, we proposed a novel feature to identify the Bacillus based on their codon usage bias, called relative synonymous codon pair usage (RSCPU). We extracted this feature from genes collected from National Center for Biotechnology Information (NCBI) website; then, K -means clustering and Support Vector Machine were applied to classify genes vectored. Finally, we used this method for Bacillus identification and obtained a result that our accuracy is about 3 times (2.93) higher than past research [1].