Improving multi-trait genomic prediction by incorporating local genetic correlations

Area of Science:

• Genomics
• Quantitative Genetics
• Bioinformatics

Background:

• Genomic prediction is crucial for precision medicine and genetic improvement in various species.
• Conventional multi-trait models rely on global genetic correlations, potentially overlooking regional variations.
• Local genetic correlation (LGC) estimation enables analysis of genetic correlations within specific genomic regions.

Purpose of the Study:

• To develop and evaluate novel multi-trait genomic prediction models that incorporate local genetic correlations (LGCs).
• To assess the impact of LGCs on prediction accuracy across different species and genomic regions.
• To compare the performance of LGC-based models against conventional multi-trait genomic best linear unbiased prediction (MTGBLUP).

Main Methods:

• Proposed three novel multi-trait genomic prediction models integrating LGCs.
• Evaluated model performance using simulated datasets.
• Validated models on three real-world datasets from human, cow, and pig populations.

Main Results:

• Local genetic correlations (LGCs) exhibit significant heterogeneity across the genome.
• Incorporating LGCs into multi-trait prediction models improved prediction accuracy by an average of 12.76% ± 2.07% compared to MTGBLUP in real datasets.
• The proposed models demonstrate enhanced predictive ability across diverse populations.

Conclusions:

• Local genetic correlations are important factors in multi-trait genomic prediction.
• Integrating LGCs into prediction models offers a substantial improvement over traditional methods.
• This approach holds promise for advancing precision medicine and genetic improvement strategies.