Genotyping errors can affect indirect predictions of young selection candidates: A simulation study

This study aimed to evaluate the effect of genotyping errors on indirect predictions (IP). For that, we simulated a dairy cattle population with 5 replicates, using QMSim. Benchmark GEBV for young animals (n = 11,250) were computed using a full single-step genomic BLUP (generations 11-15). Three SNP error rates (5%, 10%, and 20%) were applied to the genotypes of selection candidates (generation 15). A reduced analysis with generations 11 to 14 (44,500 animals) was conducted to obtain SNP solutions for computing IP in young animals using either the correct genotype files or files with genotyping errors. The average IP values decreased as the error rate increased. The correlation between the benchmark GEBV and the IP obtained with the correct genotype was 0.98 ± 0.01. As expected, this correlation decreased with increasing genotyping errors, reaching its lowest value (0.93) at 20% errors. The regression coefficient of GEBV on IP increased (i.e., more underdispersed breeding values) as the error rate increased: b₁ moved away from 1.01 ± 0.02 (no errors) to 1.33 ± 0.03 (20% errors). Results of this study highlight the importance of accurate genotype calls to achieve precise and unbiased IP when performing IP for young animals with genotypes.