Comparison of significant single nucleotide polymorphisms selections in GWAS for complex traits

Area of Science:

• Quantitative genetics
• Genomic selection
• Statistical genetics

Background:

• Accurate selection of significant single nucleotide polymorphisms (SNPs) is crucial for understanding complex traits.
• Various statistical models exist for SNP selection, each with different assumptions and capabilities.

Purpose of the Study:

• To compare the performance of four distinct SNP selection approaches (M1, M2, M3, M4) for complex traits.
• To evaluate how different models handle linkage disequilibrium (LD) and identify SNPs with varying effect sizes.

Main Methods:

• Tested 46,267 SNPs in 2601 Holstein Friesian bulls across four traits: milk yield (MY), fat yield (FY), somatic cell score (SCS), and non-return rate for heifers (NRH).
• Employed four models: single SNP (M1), single SNP with random polygenic effect (M2), non-parametric CAR score (M3), and SNP-BLUP (M4).

Main Results:

• Model performance varied, with M1 selecting numerous SNPs (except for NRH), while M2 and M3 selected fewer SNPs.
• M4 selected more SNPs than M2 and M3, and showed better handling of LD between SNPs for MY and FY.
• Functional annotation of M4-selected SNPs aligned well with known functional information and QTL mapping results.

Conclusions:

• Multiple-SNP models, such as SNP-BLUP (M4), are recommended for identifying SNPs with both strong and moderate effects on complex traits.
• These models effectively account for linkage disequilibrium, improving SNP selection accuracy.
• M4 demonstrated strong concordance with existing genetic and functional data.