Comparison of single-trait and multiple-trait genomic prediction models

Area of Science:

• Animal breeding and genetics
• Quantitative genetics
• Genomic prediction

Background:

• Comparing single-trait genomic models (STGM) and multiple-trait genomic models (MTGM) for genomic prediction.
• Simulating three scenarios: no missing data, 90% missing data in high heritability trait, and 90% missing data in low heritability trait.
• Utilizing simulated genome with 5000 SNPs, 300 QTLs, and genetic/residual correlations between two traits.

Purpose of the Study:

• To evaluate the performance of STGM versus MTGM for genomic prediction.
• To assess the impact of missing data on the reliability of genomic predictions.
• To determine the optimal model for genomic prediction under various data scenarios.

Main Methods:

• Genomic prediction using STGM and MTGM with conventional estimated breeding values (EBVs) as response variables.
• Simulation of genomic data with varying heritability and missingness.
• Analysis of prediction reliability for two traits with different heritability levels.

Main Results:

• MTGM outperformed STGM for the low heritability trait (Trait II) when no data was missing.
• MTGM showed significantly better performance than STGM when missing records were present for one trait.
• The difference in reliability was small for traits without missing data, but MTGM provided more reliable predictions for traits with missing data.

Conclusions:

• MTGM is superior to STGM for traits with low heritability and limited records.
• MTGM consistently yields more reliable genomic predictions than STGM, particularly when data is incomplete.
• The use of MTGM enhances the accuracy of genomic selection in breeding programs.