COSIGT: population-scalable genotyping of complex loci from low-coverage sequencing data using pangenome graphs.

Matching read-depth distributions to haplotype paths in a pangenome graph using cosine similarity

COSIGT assigns diploid genotypes at complex loci by matching read-depth distributions to haplotype paths in a pangenome graph using cosine similarity. Because that metric evaluates relative coverage profiles rather than absolute read counts, the tool substantially outperforms existing likelihood-based tools at low coverage (1-2X), including on degraded samples such as ancient DNA. The authors demonstrate scalability to thousands of modern and ancient genomes, enabling population-scale analyses of complex variation directly from low-coverage datasets.

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

Genotyping complex loci — segmental duplications, highly polymorphic regions — is a genuine diagnostic problem, and the pangenome graph approach is the right direction. But the demonstration explicitly targets population genetics and ancient DNA at 1-2X, whereas a diagnostic genome is sequenced around 30X: the tool's specific gain in that regime is not documented here, and the abstract reports neither per-locus sensitivity nor specificity. Worth watching for loci that standard callers leave unusable, not yet ready for a reporting pipeline.

Analysis by Dr Thibaut Benquey