Fast and sensitive classification of short metagenomic reads with SKraken

Abstract

The major problem when analyzing a metagenomic sample is to taxonomically annotate its reads in order to identify the species and their relative abundances. Many tools have been developed recently, however they are not always adequate for the increasing database volume. In this paper we propose an efficient method, called SKraken, that combines taxonomic tree and k-mers frequency counting. SKraken extracts the most representative k-mers for each species and filter out less representative ones. SKraken is inspired by Kraken, which is one of the state-of-art methods. We compare the performance of SKraken with Kraken on both real and synthetic datasets, and it exhibits a higher classification precision and a faster processing speed. Availability: https://bitbucket.org/marchiori_dev/skraken.