Taxonomic Classification at the Strain Level using a Species-of-Interest $\boldsymbol{k}$-mer Database

Metagenomic shotgun sequencing of microbial environments provides researchers with large volumes of genetic data that can elucidate the complex profile of microbial species in a microbiome. However, the ability to classify these sequences at the strain level is computationally challenging due to a high level of genetic similarity among strains. Furthermore, simply matching a read to a higher taxonomic rank (phylum, genus, species) may provide little to no clinical insight. We introduce an algorithm that classifies metagenomic reads at the strain level using exact matches of $\boldsymbol{k}$-mers against a database of species and strain-level $\boldsymbol{k}$-mers. Comparison of our method to the state of the art, Kraken, using simulated reads shows significant improvement in strain-level sensitivity and precision in the presence of sequencing error rate.