Trend of bias in prediction of genomic estimated breeding values due to selective genotyping in genomic selection schemes in consecutive generations

The aim of this study was to investigate the trend of bias in genomic estimated breeding values (GEBVs) arising from selective genotyping of the candidate population in an ongoing selection scheme. The bias was calculated as the regression of true breeding values (TBVs) on GEBVs. A simulation study was performed under two scenarios with selection intensities (SI) of 0.798 and 1.755 for three traits with heritability (h2) of 0.1, 0.25 and 0.4 in 10 consecutive generations. Regression of TBVs on GEBVs was close to one for the first generation when selective genotyping was random, and it continuously receded from one as selection shifted to choose animals with high EBVs from generations 2 to 10. Biasedness became larger with increased SI and decreased h2. Further, biasedness increased over the generations but the rate of change in biasedness decreased dramatically after the second generation and became almost steady after generation 4 which may be due to Bulmer effect. The findings showed that scaling down the GEBVs, using a scale parameter, might help removing biasedness in generation 4 onwards.

Keywords:

genomic selection , Selective genotyping , bias