Multi-PGS enhances polygenic prediction by combining 937 polygenic scores.

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Polygenic scores (PGS) are crucial for genetic prediction but require large training datasets.
• Increasing sample sizes for specific phenotypes is costly and time-consuming.
• Genetically correlated phenotypes offer a way to augment training data.

Purpose of the Study:

• To develop and validate a framework for generating multi-polygenic scores (multi-PGS).
• To improve prediction accuracy for complex traits by leveraging publicly available genome-wide association studies (GWAS).
• To enable PGS generation for phenotypes lacking existing GWAS and for case-case predictions.

Main Methods:

• A novel framework to automatically generate multi-PGS from numerous public GWAS datasets.
• Utilizing genetically correlated phenotypes to increase effective sample size for PGS training.
• Benchmarking the multi-PGS framework against existing prediction methods.

Main Results:

• The multi-PGS framework significantly increased prediction accuracy across all tested psychiatric disorders and other outcomes.
• Achieved up to a 9-fold increase in prediction R-squared for attention-deficit/hyperactivity disorder compared to single PGS.
• Successfully generated multi-PGS for phenotypes without prior GWAS and for case-case prediction scenarios.

Conclusions:

• The proposed multi-PGS framework offers a powerful and efficient method to boost genetic prediction accuracy.
• This approach overcomes limitations of single PGS by leveraging diverse GWAS data.
• The framework has broad potential applications in emerging biobanks and genetic research.