Comparison of different response variables in genomic prediction using GBLUP and ssGBLUP methods in Iranian Holstein cattle.

Autor: Afrazandeh M; Department of Animal Science, Faculty of Agriculture Sciences and Food Industries, Science and Research Branch, Islamic Azad University, Tehran, Iran., Abdolahi-Arpanahi R; Department of Animal and Dairy Science, College of Agricultural and Environmental Sciences, University of Georgia, Athens, USA., Abbasi MA; Animal Science Research Institute of Iran, Agricultural Research, Education and Extension Organization (AREEO), Karaj, Iran., Kashan NEJ; Department of Animal Science, Faculty of Agriculture Sciences and Food Industries, Science and Research Branch, Islamic Azad University, Tehran, Iran., Torshizi RV; Department of Animal Science, Faculty of Agriculture, Tarbiat Modares University, Tehran, Iran.
Jazyk: angličtina
Zdroj: The Journal of dairy research [J Dairy Res] 2022 May 23, pp. 1-7. Date of Electronic Publication: 2022 May 23.
DOI: 10.1017/S0022029922000395
Abstrakt: We compared the reliability and bias of genomic evaluation of Holstein bulls for milk, fat, and protein yield with two methods of genomic best linear unbiased prediction (GBLUP) and single-step GBLUP (ssGBLUP). Four response variables of estimated breeding value (EBV), daughter yield deviation (DYD), de-regressed proofs based on Garrick (DRPGR) and VanRaden (DRPVR) were used as dependent variables. The effects of three weighting methods for diagonal elements of the incidence matrix associated with residuals were also explored. The reliability and the absolute deviation from 1 of the regression coefficient of the response variable on genomic prediction (Dev) using GBLUP and ssGBLUP methods were estimated in the validation population. In the ssGBLUP method, the genomic prediction reliability and Dev from un-weighted DRPGR method for milk yield were 0.44 and 0.002, respectively. In the GBLUP method, the corresponding measurements from un-weighted EBV for fat were 0.52 and 0.008, respectively. Moreover, the un-weighted DRPGR performed well in ssGBLUP with fat yield values for reliability and Dev of 0.49 and 0.001, respectively, compared to equivalent protein yield values of 0.38 and 0.056, respectively. In general, the results from ssGBLUP of the un-weighted DRPGR for milk and fat yield and weighted DRPGR for protein yield outperformed other models. The average reliability of genomic predictions for three traits from ssGBLUP was 0.39 which was 0.98% higher than the average reliability from GBLUP. Likewise, the Dev of genomic predictions was lower in ssGBLUP than GBLUP. The average Dev of predictions for three traits from ssGBLUP and GBLUP were 0.110 and 0.144, respectively. In conclusion, genomic prediction using ssGBLUP outperformed GBLUP both in terms of reliability and bias.
Databáze: MEDLINE