Signal Based Feature Selection for Fast Classification of Sequences in Metagenomics

Abstract

The rapid development in DNA sequencing techniques brings completely new possibilities into metagenomics research. No longer is the whole metagenome sequencing an issue. On the other hand, this progress lies new demands on bioinformatics tools indented to process this kind of data. Unlike the amplicon based sequencing where every sequence represents a particular gene, the whole metagenome sequencing produce sequences that are random pieces of genomes in the metagenome. Therefore, the reference database for identification of these sequences cannot be used. Here, we present fast feature selection based on genomic signal processing for alignment-free classification of sequences in the metagenome.