A computationally feasible multi-trait single-step genomic prediction model with trait-specific marker weights.

Area of Science:

• Animal breeding and genetics
• Quantitative genetics
• Genomic evaluation

Background:

• Genomic evaluation models can improve prediction accuracy by accounting for varying marker influences.
• Standard multi-trait models become computationally intensive with trait-specific marker weights.

Purpose of the Study:

• To develop and implement a computationally feasible multi-trait single-step SNPBLUP model for large genomic datasets.
• To enable the use of precomputed trait-specific marker weights within the single-step framework.

Main Methods:

• Developed a modified multi-trait single-step SNPBLUP model incorporating precomputed trait-specific marker weights.
• Tested the model using simulated data and marker weights derived from BayesA.
• Evaluated computational performance (memory, time) and prediction accuracy compared to the standard model.

Main Results:

• The modified model showed slightly increased memory and per-iteration computing time.
• Model convergence was slower, leading to longer total computation time.
• Despite computational trade-offs, prediction accuracy was improved by using marker weights.

Conclusions:

• The marker-weighted single-step SNPBLUP model effectively accommodates trait-specific marker weights.
• This approach enhances prediction accuracy in genomic evaluations.
• The model is suitable for large genomic data evaluations utilizing precomputed marker weights.