14
Improving accuracies of genomic predictions by enriching 50K genotypes with markers from 770K genotypes at QTL regions
More accurate genomic predictions were expected using high-density marker panels like Illumina BovineHD BeadChip (770K) rather than Illumina BovineSNP50 BeadChips (50K) due to greater linkage disequilibrium between markers and quantitative trait loci (QTL). Results from field data showed little advantage for 770K panels in dairy cattle populations. We compared accuracies of genomic predictions for birth, weaning and yearling weights in Hereford cattle using 50K, imputed 770K, or enriched 50K genotypes (50K genotypes plus imputed 770K genotypes at locations of the largest QTL). A total of 2,980 animals were genotyped with the 50K panel. For these animals, genotypes for about 770,000 markers were imputed using BEAGLE software from 136 Irish and 228 US Hereford cattle genotyped with 770K. Only those markers imputed 770K genotypes located at 106Mb on BTA5, 38Mb on BTA6, 93Mb on BTA7 and 4Mb on BTA20, and their two 1Mb flanking windows (USDA_AIPL assembly) were added to 50K genotypes to make enriched 50K genotypes (3,004 additional markers). Six-fold cross-validation was performed using five groups for training and the sixth group for validation using either 50K, imputed 770K, or enriched 50K genotypes. Deregressed estimated breeding values were used as observations in a weighted analysis that estimated marker effects to derive molecular breeding values (MBV). Bivariate animal models were used for each trait to estimate the genetic correlation between trait and MBV as a measurement of the accuracy of genomic prediction. The accuracies of MBV for birth, weaning and yearling weights were 0.670, 0.527 and 0.605 using 50K genotypes; 0.671, 0.526 and 0.590 using 770K genotypes; 0.670, 0.550 and 0.619 using enriched genotypes, respectively. These correlations are equivalent to proportionate increases in the additive genetic variance explained for these traits of 0%, 9% and 5% using enriched 50K genotypes, respectively. These results show that the accuracies of genomic predictions can be increased for some traits by using just those markers from higher density genotypes at QTL regions.
Keywords:
Genomic breeding values, Genomic Selection, Accuracy.