MetaGS: an accurate method to impute and combine SNP effects across populations using summary statistics

Area of Science:

• Genomics
• Statistical Genetics
• Animal Breeding

Background:

• Meta-analysis combines study results to boost statistical power.
• Data sharing restrictions in genomic prediction limit reference population size.
• A practical meta-analysis method is needed for industries with privacy concerns.

Purpose of the Study:

• Develop a meta-analysis method (MetaGS) to replicate multi-trait best linear unbiased prediction (mBLUP) without raw data.
• Improve single nucleotide polymorphism (SNP) effect estimations by exploiting inter-population correlations.
• Address the challenge of differing genetic variants across populations in meta-analysis.

Main Methods:

• Developed MetaGS, a meta-analysis approach that uses summary statistics.
• Exploited correlations among populations to enhance population-specific SNP effects.
• Implemented a novel method to impute missing summary statistics without raw data.

Main Results:

• MetaGS accurately reproduced mBLUP results for milk, fat, and protein yield in Holstein and Jersey cattle.
• The method improved SNP effect estimations by utilizing relationships between populations.
• Imputing over 70% of missing SNPs achieved high accuracy (r > 0.9) with minimal impact on prediction accuracy.

Conclusions:

• MetaGS serves as a viable alternative to mBLUP when raw data sharing is not feasible.
• The method facilitates more flexible collaborations in genomic prediction.
• MetaGS offers advantages over single-trait BLUP models by enabling multi-trait analysis without data sharing.