Metagenomic Phylogenetic Classification Using Improved Naieve Bayes

摘要

The phylogenetic classification of DNA fragments from a variety of microorganisms is often performed in metagenomic analysis to understand the taxonomic composition of microbial communities. A faster method for taxonomic classification based on metagenomic reads is required with the improvement of DNA sequencer's throughput in recent years. In this research we focus on naieve Bayes, which can quickly classify organisms with sufficient accuracy, and we have developed an acceptably fast, yet more accurate classification method using improved naieve Bayes, Weightily Averaged One-Dependence Estimators (WAODE). Additionally, we accelerated WAODE classification by introducing a cutoff for the mutual information content, and achieved a 20 times faster classification speed while keeping comparable prediction accuracy.