SAIGE-QTL

SAIGE-QTL is an R package developed with Rcpp for scalable and accurate expression quantitative trait locus (eQTL) mapping in single-cell studies.

Key Features

The SAIGE-QTL method provides several advantages for single-cell eQTL analysis:

Complex data modeling: Handles repeated and complex data structures arising from multiple cells per individual and relatedness between individuals
Discrete count modeling: Specifically designed for discrete read count data typical in single-cell RNA sequencing
Scalability: Fast and scalable for large datasets - can analyze:
- 20,000+ genes
- Tens to hundreds of cell types
- Millions of cells
- Millions of genetic variants
Rare variant testing: Includes set-based tests for rare variation effects where single-variant tests are underpowered

Supported Input Formats

SAIGE-QTL accepts genotype files in multiple standard formats:

PLINK (bed, bim, fam)
BGEN
VCF
BCF
SAV

Quick Start

Install SAIGE-QTL - Choose from multiple installation methods
Review the workflow overview - Understand the analysis pipeline
Run Step 1 - Fit the null Poisson mixed model
Perform association tests - Execute cis-eQTL or genome-wide analyses

Important Note on Normalization

⚠️ Critical: Accounting for total read counts per cell is essential in single-cell eQTL mapping. Consider using:

SCTransform function from the Seurat R package, OR
Including log(total read counts) and percentage of mitochondrial read counts as an offset in the Step 1 null model

What’s New

Version 0.3.5 (February 2, 2026):

Added --solverMethod option allow user-specified step 1 null model fitting using either Sherman-Morrison-Woodbury approach (“smw” option) or the original preconditioned conjugate gradient approach (“pcg” option). For data sets with unrelated donors, for which the sparse GRM becomes the identity matrix, SMW further reduces to closed-form block-wise operations with optimal complexity. The default is to automatically detect given phenotype file, if there are multiple cells data for 1 individual and no GRM is provided, it will automatically use SMW approach.

July 31, 2025:

Export extdata/ in the docker container so users do not need to download the git repo when using the docker image wzhou88/saigeqtl:latest on example data

Version 0.3.2 (July 28, 2025):

Added --offsetCol option for using log of total read counts per cell as an offset
Enhanced installation process using pixi
See Installation logs for complete bug fixes and updates

Citation

SAIGE-QTL

Preprint: https://www.medrxiv.org/content/10.1101/2024.05.15.24307317v1

License

SAIGE-QTL is distributed under an MIT license.

Support

For questions about SAIGE-QTL, please contact: wzhou@broadinstitute.org

For technical issues and bug reports, please use the GitHub repository issue tracker.