ANI, Mash and Dashing equally differentiate between Klebsiella species

Species of the genus Klebsiella are among the most important multidrug resistant human pathogens, though they have been isolated from a variety of environments. Given the need for quickly and accurately classifying newly sequenced Klebsiella genomes, we compared 982 Klebsiella genomes using different species-delimiting measures: Average Nucleotide Identity (ANI), which is becoming a standard for species delimitation, as well as Mash, Dashing, and DNA compositional signatures, which can be run in a fraction of the time required to run ANI. ROC analyses showed equal quality in species delimitation for ANI, Mash and Dashing (AUC: 0.99), followed by DNA signatures (AUC: 0.96). The groups obtained at optimal cutoffs were largely in agreement with species designation. Using optimized cutoffs, we obtained 17 species-level groups using either ANI, Mash, or Dashing, all containing the same genomes, unlike DNA signatures which broke the dataset into 38 groups. Further use of Mash to map species after adding draft genomes to the dataset also showed excellent results (AUC: 0.99), producing a total of 28 Klebsiella species in the publicly available genome collection. The ecological niches of Klebsiella strains were found to neither be related to species delimitation, nor to protein functional content, suggesting that a single Klebsiella species can have a wide repertoire of ecological functions.