Genomic predictions for yield traits in US Holsteins with unknown parent groups

Area of Science:

• Animal Breeding and Genetics
• Quantitative Genetics
• Genomic Selection

Background:

• Traditional BLUP (Best Linear Unbiased Prediction) methods face challenges with large datasets and complex relationships.
• Genomic selection aims to improve accuracy and efficiency in animal breeding through genomic estimated breeding values (GEBV).
• Unknown parent groups (UPG) and data truncation can impact the reliability of breeding value estimations.

Purpose of the Study:

• To compare the reliability and bias of traditional BLUP, ssGBLUP with UPG for A only (SS_UPG), and ssGBLUP with UPG for both A and A22 (SS_UPG2).
• To evaluate the impact of phenotype-pedigree data truncation on GEBV accuracy and bias.
• To assess computational efficiency of different ssGBLUP models.

Main Methods:

• Utilized 6 large Holstein datasets with 80 million records and 31 million cows.
• Implemented phenotype-pedigree truncation scenarios (pre-1990, pre-2000, 2-3 ancestral generations).
• Calculated reliability and bias using deregressed proofs for bulls and predictivity for cows.

Main Results:

• SS_UPG2 achieved higher reliabilities (0.69-0.73) and unbiased predictions (regression coefficient of 1.00 ± 0.03) compared to BLUP and SS_UPG.
• SS_UPG2 showed minimal impact from data truncation on reliability and bias.
• SS_UPG2 demonstrated competitive computational efficiency (15-23 hours) compared to BLUP and SS_UPG.

Conclusions:

• ssGBLUP with UPG for both A and A22 (SS_UPG2) is the most accurate and unbiased GEBV estimation method.
• SS_UPG2 is robust to phenotype-pedigree data truncation, ensuring reliable predictions for young selection candidates.
• This advanced ssGBLUP approach enhances genomic selection accuracy in large-scale dairy cattle populations.