Ok-k-mer: Selecting k-mer length ensembles for calculating genomic distance

Ok-k-mer improves phylogeny-aware genomic distance estimation by selecting weighted k-mer length ensembles, achieving a 23.9% relative improvement in Chlamydiales with k=(9,17,22).

• Why it matters: Default k-mer lengths can distort distance estimates, and accuracy can improve when multiple k values are combined, especially for distances tied to true phylogenetic relationships.
• What they did: Given a phylogenetic tree, Ok-k-mer ranks genome pairs by Jaccard distance across k values and uses progressive search to maximize Spearman correlation (rho) between weighted k-mer ranks and branch distances across eight GTDB datasets.
• The result: Across eight GTDB datasets, ensembles improved correlation versus standard default k-mers, and in Chlamydiales the k=(9,17,22) ensemble improved 23.9% over k=20 and 36.7% over 31-mers.

AI summary of the abstract · may contain errors · not medical advice · read the paper.