Accurate estimation of variance components (VCs) is needed for genomic prediction, but until now VCs are usually estimated by restricted maximum likelihood (REML) in pedigree-based animal model (P-AM). In this study, REML in P-AM and Markov chain Monte Carlo (MCMC) procedure via Gibbs sampling in single-step Bayesian regression model (SSBR) were used to investigate the consequences of VCs estimation based on different subsets of phenotypic information. Three scenarios were analyzed: (1) phenotypes only from conventional (P-AM) part of the breeding scheme; (2) phenotypes from both conventional and genomic selection (GS) parts of the breeding scheme; (3) phenotypes only from the GS part of the breeding scheme. Unbiased estimation of VCs was obtained before GS era via P-AM. Including records from GS era led to biased estimation of VCs for P-AM. SSBR allowed utilizing data from all periods and combine information from both genotyped and non-genotyped individuals, and unbiased estimated genetic variance was achieved. With the advantage of sampling remaining genetic and marker variances separately, SSBR have the potential option to achieve unbiased VCs estimation in populations under GS. Keywords: variance components, single-step, imputation
Proceedings of the World Congress on Genetics Applied to Livestock Production, Volume Methods and Tools - Models and Computing Strategies 2, , 486, 2018
|Download Full PDF||BibTEX Citation||Endnote Citation||Search the Proceedings|
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.