A scalable Bayesian functional GWAS method accounting for multivariate quantitative functional annotations with...

Area of Science:

• Genetics
• Genomics
• Statistical Genetics

Background:

• Existing genome-wide association study (GWAS) methods for variant prioritization face limitations with non-overlapped categorical annotations and computational burden.
• Scalable integration of multivariate quantitative functional annotations is needed to enhance fine-mapping and causal variant identification in GWAS.

Purpose of the Study:

• To introduce BFGWAS_QUANT, a scalable Bayesian functional GWAS method designed to incorporate multivariate quantitative functional annotations.
• To develop a scalable algorithm for joint modeling of genome-wide variants within the BFGWAS_QUANT framework.
• To evaluate the performance of BFGWAS_QUANT in improving GWAS power and quantifying annotation enrichment.

Main Methods:

• Developed BFGWAS_QUANT, a Bayesian method for functional GWAS integrating quantitative annotations.
• Implemented a scalable computation algorithm for joint modeling of genome-wide variants.
• Applied BFGWAS_QUANT to individual-level GWAS data for five Alzheimer's disease (AD) phenotypes and summary-level IGAP AD data.

Main Results:

• BFGWAS_QUANT demonstrated accurate annotation enrichment quantification and improved GWAS power in simulations.
• Histone modification annotations, particularly H3K27me3, showed higher enrichment than eQTL annotations for AD phenotypes.
• cis-eQTLs in microglia exhibited greater enrichment than bulk brain frontal cortex eQTLs for AD.
• The APOE E4 allele was identified as a significant risk factor, and BFGWAS_QUANT fine-mapped 32 variants from 1,073 genome-wide significant variants in IGAP data.
• Polygenic risk scores derived from BFGWAS_QUANT showed comparable prediction accuracy to existing methods.

Conclusions:

• BFGWAS_QUANT is a scalable and effective tool for prioritizing potential causal variants in GWAS by leveraging multivariate quantitative functional annotations.
• The method provides valuable insights into AD genetic architecture, highlighting the importance of histone modifications and microglial eQTLs.
• BFGWAS_QUANT enhances the utility of GWAS for genetic discovery and risk prediction.