Improving the Accuracy of Genomic Prediction of Milk Traits in the New Zealand Holstein Friesian Population07 May 2013
Altering a gene mutation that codes for milk traits such as fat yield has been investigated by M.Hayr, M. Saatchi, D. Johnson and D. Garrick and revealed at the mid-west American Dairy Science Association - held last month.
"Accurate Genomic Estimated Breeding Values (GEBV) are crucial for preventing the accumulation of inaccuracies when unproven parents are selected. This study investigated the effect of including the DGAT1 mutation, a major gene for milk traits, in calculating GEBV.
Data on 5,661 Holstein Friesian cows were provided by LIC, a dairy cattle breeding company in New Zealand, and included Illumina SNP50 (50k) genotypes and Deregressed Estimated Breeding Values (DEBV) for fat yield. DGAT1 genotypes were provided for 1,133 cows and DGAT1 genotypes were imputed for the remaining 4,528 cows using BEAGLE. Three models were run in GenSel using the Bayes C method and 5-fold cross-validation with 2.5 per cent of SNPs assumed to have an effect on the trait: 1) a model relying on linked 50k markers to pick up the effect of DGAT1; 2) a model with 50k markers and DGAT1 fit as random effects; and 3) a model with 50k markers as random effects and DGAT1 genotype as a fixed effect.
These models were run on all cows then repeated using only cows where DGAT1 had been directly genotyped. The GEBV accuracy was defined as the simple correlation between DEBV and GEBV, whereas the genetic correlation between GEBV and BV would be approximately twice this magnitude after dividing by root heritability to account for residual variation in GEBV.
The regression of GEBV on DEBV was obtained to quantify bias in the GEBVs. Accuracy was lowest when only 50k markers were included in the model and increased when DGAT1 was included in the model, with the highest accuracy when DGAT1 was fit as a fixed effect.
Regression coefficients were close to one to indicating little to no bias in GEBV, however the least bias was when DGAT1 was fit as a fixed effect. These results suggest that including DGAT1 genotype as a fixed effect when calculating GEBV both increases accuracy of the GEBV and reduces bias.
Table 1. Regression Coefficients (b) and Correlations (r),
Between DEBV and GEBV