Preprint warns GWAS SNP localisation may mislead in livestock under diffuse polygenic architecture

A preprint posted to bioRxiv reports that widely used genome-wide association study (GWAS) methods may be poorly suited to identifying causal variants at the individual SNP level in livestock populations characterised by diffuse polygenic architecture.

The authors constructed a benchmarking framework using real Duroc pig genotypes, simulating phenotypes under conditions where trait variation is spread across many variants of small effect rather than concentrated at discrete loci. They found that standard GWAS approaches can generate strong, apparently localised association signals in the absence of any genuine nearby causal variant. The study attributes this behaviour to features common in livestock populations: small effective population sizes, pronounced family structure, extended linkage disequilibrium (LD) across broad genomic intervals, and diffuse polygenic architecture. Under these conditions, the cumulative effects of many variants across a genomic region can be captured jointly by whichever SNP happens to be best tagged in a particular study, producing a misleadingly specific localisation signal.

The findings have practical implications for livestock genomics programmes that use GWAS hits to prioritise variants for functional follow-up, fine-mapping, or marker-assisted selection. The authors do not argue that GWAS is uninformative in livestock, but that variant-level conclusions should be interpreted with caution where the underlying architecture is likely to be highly polygenic.

This work is a preprint and has not yet completed peer review.