Metagenome Classification Using Naïve Bayes Classifier Method.

Abstract

Metagenome study is an important step in the taxonomic grouping. Grouping can be conducted using binning method. Binning is required to determine contigs of each phylogenetic species groups. In this study, binning is done using supervised learning based composition approach. We used NaïveBayes Classifier method for performing supervised learning and employed counting of k-mer frequencies for extracting feature. The classification process was conducted at genus-level taxon. The results showed that using short fragments (400 bp), our method could obtain the accuracy of 49.34 % and 53.95 % with features of 3-mers dan 4-mers frequencies, respectively. Meanwhile, the accuracy of our method was significally increased when classifying long fragments (10 kbp). Our method could obtain the accuracy of 82.23% with 3-mers frequencies feature and 85.89% with 4-mers frequencies feature. It can be concluded that the accuracy of our classifier was increased by increasing the size of fragments. Moreover, in this research, the 4- mers frequencies feature gave the best results for classifying metagenome fragments.