New Scoring Framework Ranks Viral K-mers by Biological Significance

Researchers have introduced a computational framework called the KmerSignificance Score (KSS) designed to identify which short sequence fragments, or k-mers, best distinguish closely related viral strains while also carrying biological meaning. The work, led by Dylan Lebatteux, Florian Corso, Hugo Soudeyns, Isabelle Boucoiran, Soren Gantt and Abdoulaye Baniré Diallo, addresses a gap in existing k-mer methods, which the authors say lack standardized ways to evaluate which k-mers are most informative and focus mainly on statistical discriminative power without integrating biological context.

The framework combines three components, each bounded between 0 and 1 so that positions within a dataset can be ranked on a common numerical scale. The first is an information-theoretic measure of strain-distinguishing capacity. In benchmarking, this discriminative component achieved a mean F1 score of 0.881, which was not significantly different from chi-squared at 0.876, the strongest of eight baselines tested. Notably, however, the top-ranked selections from the discriminative component differed from those produced by chi-squared, particularly for SARS-CoV-2 and HIV-1, suggesting the two approaches prioritize different sequence regions.

The second component is an optimized amino acid substitution matrix, named MIYATA_EVO, used to assess the likely impact of mutations. Optimized with a genetic algorithm, the new matrix increased a composite agreement score across eight amino acid characteristics by 28.4 percent over the original MIYATA matrix. The third component is a protein characterization score derived from annotations and database cross-references retrieved through UniProt. Across 17,471 viral proteins, this score tracked UniProt’s five-level annotation score closely, with a Kendall’s tau of 0.775, an outcome the authors describe as expected given the overlapping annotation content of the two measures. Within each annotation level, the continuous score offered finer numerical resolution among proteins.

To test whether the framework’s rankings correspond to known biology, the team performed literature cross-referencing on three viruses: SARS-CoV-2, using 278,738 sequences and 19 Pango-derived lineage labels; HIV-1, with 12,223 sequences and 15 subtype labels; and human cytomegalovirus, with 399 to 646 sequences and 4 to 8 locus-specific genotypes. In all three cases, the top-ranked k-mers mapped to established variant-defining mutations, subtype-specific polymorphisms and genotype markers.

The authors position KSS as a standardized framework for viral k-mer prioritization with potential applications in variant surveillance, molecular epidemiology and functional annotation. The tool is publicly available through the bioinfoUQAM repository on GitHub. Because the reported benchmarks are computational evaluations and literature concordance checks rather than prospective clinical studies, the framework’s practical value in surveillance workflows will depend on further validation, but the results indicate that combining statistical discrimination with mutation-impact and annotation-based measures can surface biologically meaningful sequence positions that purely statistical rankings may overlook.

Subject of Research: A computational framework for prioritizing biologically significant k-mers in viral genomes

Article Title: KmerSignificance score: A discriminative and biologically informed framework for viral k -mer prioritization

Article References: Lebatteux, D., Corso, F., Soudeyns, H., Boucoiran, I., Gantt, S., & Baniré Diallo, A. (2026). KmerSignificance score: A discriminative and biologically informed framework for viral k-mer prioritization. PLOS Computational Biology, 22(10), e1014332. https://doi.org/10.1371/journal.pcbi.1014332

Image Credits: AI Generated

DOI: 10.1371/journal.pcbi.1014332

Keywords: virology, k-mer, genomics, SARS-CoV-2, HIV-1, bioinformatics, variant surveillance, molecular epidemiology, UniProt, amino acid substitution, KmerSignificance, score