JointPRS: A Data-Adaptive Framework for Multi-Population Genetic Risk Prediction Incorporating Genetic Correlation

Area of Science:

• Population Genetics
• Statistical Genomics
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) exhibit limited genetic prediction accuracy in non-European populations due to smaller sample sizes.
• Parameter tuning for predictive models is challenging with small datasets, particularly for non-European ancestries.

Purpose of the Study:

• To introduce JointPRS, a novel, data-adaptive framework for simultaneous multi-population genetic prediction using GWAS summary statistics.
• To enhance genetic prediction accuracy in underrepresented populations by leveraging genetic correlation structures.

Main Methods:

• Developed JointPRS, a framework that models multiple populations concurrently using GWAS summary statistics.
• Incorporated genetic correlation structures into the prediction model.
• Utilized a data-adaptive approach for robust performance with limited tuning data.

Main Results:

• JointPRS demonstrated superior performance compared to six state-of-the-art methods across various data scenarios (no tuning, same-cohort tuning/testing, different-cohort tuning/testing).
• The framework achieved high accuracy for most traits in non-European populations (East Asian, African, South Asian, Admixed American).
• Evaluations included 22 quantitative and 4 binary traits using UK Biobank and All of Us data.

Conclusions:

• JointPRS offers a robust and computationally efficient solution for improving genetic prediction in diverse populations.
• The framework effectively addresses challenges posed by limited sample sizes and tuning data in non-European ancestries.
• JointPRS maintains model simplicity while enhancing prediction accuracy across multiple continental populations.