Estimation & Prediction

CNV-assisted genetic parameters and genome-wide associations for low-heritability traits

---

Abstract

Copy number variations (CNVs) can change gene and regulatory element dosages, which are not fully captured by single nucleotide polymorphisms (SNPs). Hence, CNVs might capture additive genetic variance beyond SNPs and reveal trait-associated genomic regions missed by SNP-based analyses. After mapping 4,113 CNVs from high-density SNP genotyping data of 3,601 Holsteins using PennCNV, we quantified the contribution of CNVs to additive genetic variance and conducted CNV-based and SNP-based GWAS to detect genetic variants underlying retained fetal membranes (RFM; n=3,305), metritis (MET; n=3,305), clinical endometritis (CE; n=3,303), resumption of ovarian cyclicity (ROC; n=3,303), and pregnancy status on day 60 after insemination (P60; n=3,207). For estimation of variance components, a SNP-derived genomic relationship matrix (GSNP; 578,677 SNPs) and a CNV-derived GRM (GCNV; 4,154 CNVs) were created. The variance components for MET, CE, ROC, and P60 were estimated by fitting either only the GSNP (SNP-based) or jointly GSNP and GCNV (SNP+CNV-based) into threshold models that included parity category and farm-year-season as fixed effects. The SNP+CNV-based model increased additive genetic variance estimates for all traits compared to the SNP-based model, with relative increases ranging from 26.1% for ROC to 121.9% for CE. Consequently, the SNP+CNV-based model yielded greater heritability estimates than the SNP-based model, changing from 0.071±0.025 to 0.088±0.026 (+23.9%) for MET, 0.039±0.025 to 0.084±0.026 (+115.4%) for CE, 0.081±0.025 to 0.100±0.026 (+23.5%) for ROC, and 0.090±0.025 to 0.177±0.026 (+96.7%) for P60. The RFM models failed to converge because of an imbalanced case-control proportion. For association tests, we controlled relatedness by fitting a GRM constructed from 50,000 SNPs randomly sampled in proportion to chromosome size. We tested 578,677 SNPs or 2,654 CNVs, one marker at a time, alongside a random polygenic effect and fixed effects of farm, year, season, and parity category. CNV-GWAS revealed a BTA29 duplication associated with CE and P60 (PP< 0.10), which was not detected by SNPs. This region harbors candidate genes related to uterine health in humans, sheep, and cattle. Cows carrying the BTA29 duplication showed approximately twice the prevalence of CE and MET observed in non-carriers and had about five times lower confirmed pregnancy. Additionally, a BTA22 deletion was suggestively associated with RFM. Notably, this genomic region, flagged exclusively by the CNV-GWAS, harbors candidate genes involved in epithelial cell detachment and other placental functions in humans. Furthermore, carriers of the BTA22 deletion exhibited roughly seven times greater RFM prevalence than non-carriers. No CNV was associated with ROC. Overall, CNVs captured additive genetic variance beyond SNPs alone, increasing heritability estimates, and revealed trait-associated genomic regions missed by SNP-GWAS. Moreover, CNV-informed genomic models may enhance the prediction of breeding values for low-heritability traits by explaining additive genetic variance not captured by SNPs alone.

Keywords: 2026

How to Cite:

Ladeira, G., Pinedo, P., Santos, J., Thatcher, W. & Rezende, F., (2026) “CNV-assisted genetic parameters and genome-wide associations for low-heritability traits”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2287255. doi: https://doi.org/10.31274/wcgalp.24249

Rights: 1