BFGWAS_QUANT

BFGWAS_QUANT applies a scalable Bayesian framework to integrate multivariate quantitative functional annotations into genome-wide association studies to quantify annotation enrichment and prioritize potential causal variants.

Key Features:

• Scalability and Computation Efficiency: Implements a scalable computation algorithm that enables joint modeling of genome-wide variants for large GWAS datasets.
• Multivariate Quantitative Functional Annotations: Models multivariate quantitative functional annotations instead of non-overlapping categorical annotations.
• Annotation Enrichment Quantification: Quantifies annotation enrichment through rigorous simulation studies to improve GWAS power.
• Application in Alzheimer's Disease Research: Applied to five Alzheimer's disease (AD)-related phenotypes using individual-level GWAS data from ~1,000 participants and identified H3K27me3 (polycomb regression) as having higher enrichment than eQTL annotations across phenotypes.
• Cis-eQTLs in Microglia: Revealed that cis-eQTLs in microglia exhibit higher enrichment than eQTLs from bulk brain frontal cortex tissue for the studied phenotypes.
• Validation with Summary-Level Data: Validated findings using summary-level GWAS data from IGAP (International Genomics of Alzheimer's Project) comprising ~54,000 participants.
• Fine-Mapping Capabilities: Fine-mapped 32 significant variants from 1,073 genome-wide significant variants identified in the IGAP data.
• Polygenic Risk Scores (PRS): Produces effect size estimates used to derive PRSs with prediction accuracy comparable to methods based on sparse causal models.

Scientific Applications:

• Alzheimer's Disease Genetics: Identifies enriched annotations and prioritizes candidate causal variants for AD-related phenotypes, including histone modifications and microglial cis-eQTLs.
• Annotation Enrichment Analysis: Quantifies enrichment of multivariate quantitative functional annotations to inform biological interpretation of GWAS signals.
• Fine-Mapping of GWAS Loci: Prioritizes candidate causal variants by jointly modeling genome-wide variants for fine-mapping of significant loci.
• Polygenic Risk Prediction: Generates effect size estimates for constructing and evaluating PRSs against sparse causal-model–based methods.

Methodology:

Uses a scalable Bayesian model and computation algorithm to jointly model genome-wide variants with multivariate quantitative functional annotations; quantifies annotation enrichment via simulation studies; performs fine-mapping; derives effect size estimates for PRS and validates results with individual-level and IGAP summary-level GWAS data.