Abstract

The current study was undertaken to explore the efficacy of genomic prediction using single-step genomic BLUP (ssGBLUP) with information on breed-of-origin of alleles at each marker of genotyped animals in Holstein x B. indicus crossbred (CBHF) cattle of India. Breed-of-origin was used in construction of breed-of-origin specific genomic relationship matrix (G-matrix). Performance of conventional ssGBLUP (Model-I) was compared to breed-of-origin aware genomic prediction (Model-II). Comparisons were by two schemes of five-fold cross-validation, 1: wherein half-sibs families’ (daughters’) performance records were left-out from the data sets (Data set - I & Data set - II (the latter with pedigree corrections)) and 2: wherein half-sibs families’ (daughters’) performance records of sires born of proven bulls, were left-out (Data set - III). For daughters, the criteria for comparison was the gain/ loss in correlations between corrected daughter yield (Yc) and their predicted breeding values (EBV) and for sires, the criteria for comparison was the gain/ loss in correlations between average Yc of daughters of sires and sire EBV, in Model-II compared to Model-I. Correlations for all daughters did not improve in Data sets - I and II, but improvement was noted in Data set - III. Correlations for non-genotyped daughters reduced in Data sets-I and II, but improved for Data set - III. Correlations across genotyped daughters improved consistently (around +5%) with all three data sets. Correlations across all sires did not improve in Data sets-I and II. Data sets - I and II included all data for validation schemes, so sires with fewer daughters were also included, which may be a cause of no gain being observed due to large uncertainty. However, gains were noted in all three data sets for genotyped sires (there were no non-genotyped sires in Data set - III), the biggest increase was in Data set - III (+15%). Predicted breeding values were less biased with Data set - III for validation in Model-II than in Model-I, while the reverse was the case for the other two data sets. Data set - III was consistent with practical scenarios wherein breeding values for young bulls from proven sires would be evaluated at regular intervals. Thus, using breed-of-origin specific G-matrix in genomic predictions by ssGBLUP improves the accuracy and unbiasedness of genomic estimated breeding values for CBHF genotyped animals, as compared to conventional G-matrix. This will improve genetic gains per unit of time in the Holstein x B. indicus crossbred population in India. Keywords: chromosome painting, breed of origin, ssGBLUP, HF crossbred, small holder

Swapnil Gajjar, Bernt Guldbrandtsen, Guosheng Su, Nilesh Kumar Nayee, Goutam Sahana, Kamlesh Trivedi, Mogens Lund

Proceedings of the World Congress on Genetics Applied to Livestock Production, Volume Theory to Application 3, , 754, 2018