Multivariate adaptive shrinkage improves cross-population transcriptome prediction and association studies in...

Area of Science:

• Genomics
• Statistical Genetics
• Bioinformatics

Background:

• Transcriptome prediction models trained on European-descent data exhibit reduced accuracy in other populations.
• Genetic variations, including linkage disequilibrium patterns and allele frequencies, differ across populations, impacting model performance.
• Leveraging shared regulatory effects across populations is a potential strategy to enhance cross-population transcriptome prediction.

Purpose of the Study:

• To evaluate the accuracy of different transcriptome prediction models in population-matched and cross-population scenarios.
• To assess the applicability of these models in multiethnic transcriptome-wide association studies (TWASs).
• To identify methods that improve genetic discovery in diverse populations.

Main Methods:

• Developed transcriptome prediction models using elastic net, JTI, Matrix eQTL, MASHR, and TIGAR.
• Tested out-of-sample prediction accuracy in population-matched and cross-population settings.
• Integrated models with multiethnic GWAS summary statistics from PAGE and PanUKBB for TWAS analysis.

Main Results:

• MASHR models demonstrated superior or equivalent transcriptome prediction accuracy in both population-matched and cross-population comparisons.
• MASHR models yielded a higher number of replicating discoveries in multiethnic TWASs across both PAGE and PanUKBB datasets.
• Identified both previously known and novel genetic loci through TWAS analysis using MASHR.

Conclusions:

• The study highlights the importance of methods that incorporate effect size estimates from diverse populations to improve TWAS.
• MASHR models show promise for enhancing genetic discovery in multiethnic and underrepresented populations.
• Improved cross-population prediction accuracy is crucial for equitable genomic research.