Heterogeneity-aware integrative regression for ancestry-specific association studies.

Area of Science:

• Genetics
• Bioinformatics
• Statistical genomics

Background:

• Proteome-wide association studies (PWAS) are crucial for understanding complex diseases across diverse ancestral populations.
• Accurate prediction of protein expression from genetic data (SNP genotypes) is essential for PWAS.
• Genomic studies have historically underrepresented certain ancestral groups, limiting the generalizability of findings.

Purpose of the Study:

• To develop an improved statistical method for fitting ancestry-specific joint protein quantitative trait loci (pQTL) models.
• To enhance the prediction accuracy of protein expression in ancestrally diverse populations, particularly those historically underrepresented.
• To facilitate the discovery of novel genetic associations with complex diseases in these populations.

Main Methods:

• Proposed a new penalized maximum likelihood estimator for ancestry-specific joint pQTL models.
• Developed an estimator that borrows information across ancestral groups while accommodating heterogeneity in error variances and regression coefficients.
• Introduced a convex and scale-invariant objective function parameterization and an computationally efficient approximate method.

Main Results:

• The proposed method significantly improved protein expression prediction accuracy in individuals of African ancestry.
• Downstream PWAS analyses using the enhanced predictions led to the discovery of multiple novel associations between protein expression and blood lipid traits in the African ancestry population.
• The method demonstrated improved performance in predicting protein quantitative trait loci (pQTLs) in ancestrally diverse groups.

Conclusions:

• The novel penalized maximum likelihood estimator enhances protein expression prediction accuracy in underrepresented ancestral populations.
• This improved prediction facilitates the discovery of ancestry-specific genetic associations for complex diseases, such as blood lipid traits.
• The method offers a valuable tool for advancing precision medicine across diverse global populations.