Multiple-trait genomic evaluation of linear type traits using genomic and phenotypic data in US Holsteins

Area of Science:

• Animal Genetics and Breeding
• Quantitative Genetics
• Dairy Cattle Evaluation

Background:

• Current USDA genomic evaluations for US Holsteins utilize a single-trait (ST) model with multiple intermediate steps.
• Genomic predictions are crucial for selecting superior dairy sires, but model efficiency and accuracy are ongoing research areas.
• Previous methods involved complex procedures, necessitating a more streamlined and accurate approach for evaluating multiple traits simultaneously.

Purpose of the Study:

• To implement and assess a unified single-step multiple-trait (MT) model for genomic evaluation of 18 linear type traits in US Holsteins.
• To compare the accuracy and computational efficiency of the MT model against the traditional ST model.
• To investigate the impact of different genomic relationship matrices (G) and weighting strategies on prediction accuracy.

Main Methods:

• Phenotypic type data from 4,813,726 Holsteins and SNP marker data from 17,293 bulls were used.
• Genomic predictions were computed using both ST and MT models with various genomic relationship matrices (G).
• Model performance was evaluated using coefficients of determination (R²) and regression of predictions on daughter deviations.

Main Results:

• The MT model demonstrated higher accuracy (average R² of 37%) compared to the ST model (average R² of 34%).
• Regression coefficients indicated potential bias in estimated breeding values, which could be mitigated by adjusting the weight for the numerator relationship matrix (A⁻¹).
• While MT models required more computational time per iteration (3-6 min) than ST models (5s), the single-step approach proved feasible for 18 traits.

Conclusions:

• The MT single-step approach is a feasible and more accurate method for genomic evaluation of multiple linear type traits in US Holstein cattle.
• Switching from ST to MT models significantly increases the accuracy of genomic evaluations.
• Adjusting the weight for the A⁻¹ in genotyped animals can help reduce inflation in genomic evaluations for young bulls.