Preprint: AI-assisted analysis of mouse copy number variant database identifies candidate locus for metabolic syndrome

A preprint deposited on bioRxiv describes an analysis of long-read sequencing data from 40 inbred mouse strains to produce a comprehensive catalogue of copy number variants (CNVs) — structural variants in which segments of the genome are duplicated or deleted relative to a reference sequence.

The authors used pangenome graph-based methods alongside a telomere-to-telomere (T2T) reference assembly for the C57BL/6J strain to resolve 1,594 high-confidence CNVs, including variants in repeat-rich and segmentally duplicated regions that are typically intractable to short-read approaches. An AI-assisted analytical step was applied to interrogate the resulting CNV database for associations with phenotypes relevant to the metabolic syndrome — a cluster of traits including obesity, dyslipidaemia, and insulin resistance.

The study identifies a candidate CNV-linked genetic factor in a murine metabolic syndrome model. The authors frame this as an example of how structural variation, which is underrepresented in standard SNP-based genome-wide association studies, may contribute to the missing heritability of complex traits.

The work is methodologically notable for its use of long-read sequencing and pangenome graph assembly in a multi-strain mouse resource, providing a structural variant map that may be of broad utility to researchers using inbred mouse models of human metabolic disease. This is a preprint and has not yet undergone peer review.