SAIGEgds

SAIGEgds implements scalable generalized mixed models with highly optimized C++ code and Genomic Data Structure (GDS) file integration for single variant tests and set-based aggregate tests in large-scale Phenome-wide Association Studies (PheWAS).

Key Features:

• Highly optimized C++ implementation: Enables analysis of millions of variants and samples with significantly faster performance than the SAIGE R package.
• GDS file integration: Directly reads and processes Genomic Data Structure files for efficient data management.
• Single variant tests: Performs association testing for individual genetic variants.
• Set-based aggregate tests: Includes state-of-the-art ACAT-O set-based tests for gene-level or region-level analysis.
• Sample structure control: Accounts for relatedness and population structure in association analyses.
• Case-control imbalance handling: Adjusts for unequal numbers of cases and controls in binary trait analyses.
• Optional GPU acceleration: Supports OpenCL-based GPU acceleration for GRM cross-product computation in null model fitting and for GRM construction.

Scientific Applications:

• Large-scale Phenome-wide Association Studies (PheWAS): Enables systematic testing of associations between millions of genetic variants and a wide range of phenotypes.
• Genetic association studies with sample structure: Analyzes data from cohorts with related individuals or population stratification.
• Imbalanced case-control studies: Analyzes binary traits with unequal case and control proportions, such as rare disease studies.

Methodology:

Based on the SAIGE R package (v0.45, Zhou et al. 2018 and Zhou et al. 2020) with extensions for ACAT-O set-based tests. Uses highly optimized C++ implementation with GDS file integration. Optional OpenCL-based GPU acceleration for GRM cross-product computation in null model fitting and GRM construction.