Genomic-polygenic and polygenic evaluations for milk yield and fat percentage using random regression models with Legendre polynomials in a Thai multibreed dairy population

• We estimated genetic parameters using phenotypes, pedigree, and genomic information.
• We imputed animals genotyped with GeneSeek GGP9k, GGP20k, and GGP26k to GGP80k.
• We compared random regression genomic and polygenic models for milk yield and fat%.
• Random regression genomic heritabilities were higher than polygenic for both traits.
• EBV accuracies were higher for random regression genomic than polygenic models.

The objectives of this research were to compare estimates of variance components, genetic parameters, prediction accuracies, and rankings of animals for 305-d milk yield (305-d MY) and 305-d fat percentage (305-d FP) from random regression genomic-polygenic (RRGM) and random regression polygenic (RRPM) models. In addition, RRGM and RRPM prediction accuracies and rankings were compared with those from a standard cumulative 305-d genomic-polygenic model (SCGM). The dataset contained first-lactation monthly test-day records (69,029 for MY and 29,878 for FY) from 7206 Holstein-upgraded cows located in 761 Thai farms. Genotypic data included 74,144 actual and imputed SNP from 1661 animals. Variance components and genetic parameters were estimated using REML procedures. The RRGM and RRPM included contemporary group (herd-year-season), calving age, heterosis, and third-order Legendre population regression coefficients. Random effects were animal additive genetic third-order Legendre regression coefficients, permanent environment third-order Legendre regression coefficients, and residual. The SCGM contained contemporary group (herd-year-season), calving age and heterosis as fixed effects, and additive genetic and residual as random effects. The RRGM yielded higher additive genetic variances and heritabilities for 305-d MY and 305-d FP than RRPM, whereas correlations between MY and FY were similar in both models. The highest prediction accuracies for both traits were for RRGM, followed by RRPM, and the lowest ones were from SCGM. Similarly, the highest rank correlations were between animal EBV for 305-d MY and 305-d FP from RRGM and RRPM, followed by those between RRGM and SCGM, and the lowest ones were between RRPM and SCGM. The higher heritability estimates and higher prediction accuracies for RRGM than for RRPM and SCGM indicated that higher selection responses for 305-d MY and 305-d FP may be achieved in this Thai dairy population by utilizing a random-regression model and genotypic information in addition to phenotypes and pedigree.