TorchGWAS 1.0: GPU-accelerated GWAS at scale.
- 1Department of Bioinformatics and Systems Medicine, McWilliams School of Biomedical Informatics, The University of Texas Health Science Center at Houston, Houston, TX 77030, United States.
- 2Department of Biomedical Informatics and Data Science, Yale School of Medicine, New Haven, CT 06510, United States.
Abstract
Motivation
Imaging, molecular and machine-learning workflows can generate thousands of quantitative phenotypes in one cohort. Repeating genome-wide scans trait by trait creates a computational and output bottleneck.Results
TorchGWAS is a GPU-accelerated framework that uses batched operations for high-throughput, covariate-adjusted linear association testing across large panels of quantitative phenotypes. Across 500 036 allele-harmonized tests, its test statistics agreed with PLINK 2.0, with a Pearson correlation difference from one of 1.3 × 10⁻ ¹ ². On an NVIDIA H100 with 80 gigabytes of memory and a 48-core Intel Xeon Gold 6442Y host, median end-to-end runtimes for 4.57 billion associations spanning 8 931 083 variants, 512 phenotypes and 35 365 samples ranged from 28.46 to 58.95 seconds across four genotype formats. These measurements included writing 36.7 gigabytes of binary summary statistics. TorchGWAS supports efficient fixed-effect association screening at scale.Availability and implementation
TorchGWAS is implemented in Python and distributed as a documented source repository at https://github.com/ZhiGroup/TorchGWAS.
Related
- Conditional-prevalence correction for liability-scale analyses of covariate-adjusted binary GWAS
- Large-scale GWAS integrated with image-based phenotyping reveals loci and trait-associated markers for Perilla frutescens seed traits.
- Copy number variants (CNVs) interfere with human accelerated regions (HARs) in neurodevelopmental disorders: a large-scale cohort study
- flashscenic: Accelerated GRN Analysis with RegDiffusion for Large-Scale Single-Cell Data
- GWAS of Leaf Morphological and Chlorophyll-Related Traits Provides Genetic Resources for Improving Photosynthetic Traits in Sugarcane
BibTeX
@article{zhao2026torchgwas,
title = {{TorchGWAS 1.0: GPU-accelerated GWAS at scale.}},
author = {Xingzhong Zhao and Ziqian Xie and Sheikh Muhammad Saiful Islam and Tian Xia and Cheng Chen and Degui Zhi},
journal = {Bioinformatics advances},
year = {2026},
month = oct,
doi = {10.1093/bioadv/vbag310},
url = {https://europepmc.org/article/MED/42910066},
note = {CC BY},
}
RIS
TY - JOUR
TI - TorchGWAS 1.0: GPU-accelerated GWAS at scale.
AU - Xingzhong Zhao
AU - Ziqian Xie
AU - Sheikh Muhammad Saiful Islam
AU - Tian Xia
AU - Cheng Chen
AU - Degui Zhi
PY - 2026
DA - 2026/10/13
JO - Bioinformatics advances
DO - 10.1093/bioadv/vbag310
UR - https://europepmc.org/article/MED/42910066
AB - Motivation Imaging, molecular and machine-learning workflows can generate thousands of quantitative phenotypes in one cohort. Repeating genome-wide scans trait by trait creates a computational and output bottleneck.Results TorchGWAS is a GPU-accelerated framework that uses batched operations for high-throughput, covariate-adjusted linear association testing across large panels of quantitative phenotypes. Across 500 036 allele-harmonized tests, its test statistics agreed with PLINK 2.0, with a Pearson correlation difference from one of 1.3 × 10⁻ ¹ ². On an NVIDIA H100 with 80 gigabytes of memory and a 48-core Intel Xeon Gold 6442Y host, median end-to-end runtimes for 4.57 billion associations spanning 8 931 083 variants, 512 phenotypes and 35 365 samples ranged from 28.46 to 58.95 seconds across four genotype formats. These measurements included writing 36.7 gigabytes of binary summary statistics. TorchGWAS supports efficient fixed-effect association screening at scale.Availability and implementation TorchGWAS is implemented in Python and distributed as a documented source repository at https://github.com/ZhiGroup/TorchGWAS.
SN - 2635-0041
ER -