TIGAR-V2

TIGAR-V2 performs transcriptome-wide association studies (TWAS) by training and applying gene expression imputation models to map genetic regulation of gene expression and identify risk genes for complex traits and diseases.

Key Features:

• Direct VCF File Reading: Reads Variant Call Format (VCF) files directly for genotype input.
• Parallel Computation Support: Supports parallel computation to accelerate large-scale analyses.
• Reduced Computational Cost: Reduces computational resource requirements—reported up to 90% reduction compared to its predecessor—through optimized genotype data loading.
• Flexible Gene Expression Prediction Models: Trains gene expression imputation models using nonparametric Bayesian Dirichlet Process Regression (DPR) or Elastic-Net.
• Support for Various GWAS Data Types: Accepts both individual-level and summary-level Genome-Wide Association Study (GWAS) data.
• Comprehensive Statistical Approaches: Implements both burden and variance-component statistics for gene-based association tests.
• Bayesian cis-eQTL Weights: Computes and leverages Bayesian cis-eQTL weights for gene expression prediction.
• Integration with GTEx V8 LD: Integrates linkage disequilibrium (LD) information from GTEx V8.

Scientific Applications:

• Cross-tissue Model Training: Trains gene expression prediction models using DPR across 49 tissue types from the GTEx V8 dataset to map transcriptional regulation.
• Cancer TWAS: Applied to TWAS of breast and ovarian cancer using public GWAS summary statistics, identifying 88 breast-cancer risk genes and 37 ovarian-cancer risk genes, with ~95% proximal or known from GWAS and ~40% overlapping previous TWAS, and reporting three novel independent TWAS risk genes implicated in carcinogenesis.

Methodology:

Directly reads VCF files, uses parallel computation and optimized genotype data loading, trains expression imputation models with nonparametric Bayesian Dirichlet Process Regression (DPR) or Elastic-Net, computes Bayesian cis-eQTL weights, integrates LD from GTEx V8, and performs burden and variance-component gene-based association tests.