Abstract
Accurate microbial isolate identification and confirmation of culture purity are fundamental prerequisites for generating reproducible research outcomes and ensuring the long-term reliability of microbial collections. Genomic data provide an ideal means of achieving high-quality and robust results concerning isolate taxonomic identification and purity assessment, particularly when other analytical methods cannot identify an organism. However, analyses based on assembled sequences, whether through marker genes or ANI, may overlook contamination present only in unassembled reads or discarded contigs. Here, we present Karmii, a Kraken 2-based workflow designed to confirm species identification and purity of sequenced isolates from short- or long-read data. The specificity of the method is to query single-isolate reads in a two-step process, first against a generic database to generate a shortlist of candidate species and second towards the pre-selected single-species databases. Having reads classified against only one species would strongly support purity and robust identification. Results are summarized in an intuitive plot showing the proportion of classified reads across all tested species as classification stringency increases. Thresholds are used to guide decision-making. To obtain reliable results, the composition of the reference database is crucial and must maximize represented diversity while excluding inconsistent taxonomic annotations. We therefore also describe a procedure for building a resource-efficient, comprehensive and curated reference database to support the analysis. A Nextflow implementation of this workflow for prokaryotic isolate analysis is provided with this article.

This publication has 18 references indexed in Scilit: