GCTAg: scalable mixed-model analysis for biobank-scale agricultural cohorts
Abstract
Genome-wide association studies identify genomic variants associated with traits. Mixed linear model association (MLMA) methods using a whole-genome relationship matrix, such as those implemented in GCTA, are powerful but computationally expensive. Here, we remove key memory and CPU bottlenecks in GCTA, reducing REML memory usage by nearly 75% and substantially accelerating MLMA by orders of magnitude in biobank-scale cohorts while preserving exactness. We further exploit relatedness in the mapping cohort through a reduced-rank Woodbury matrix approach, delivering further orders of magnitude performance gains with controlled genomic inflation. Native on-the-fly dominance recoding also eliminates slow I/O-operations on intermediate files, enabling efficient additive and dominance MLMA analyses in large agricultural cohorts.
Related
- GROMTools: scalable individual-level GReX imputation for mega-biobank-scale cohorts
- Scalable Agricultural Microbiome Sampling: Operational Definitions, Pooling Strategies, and Preservation Methods
- An updated roadmap for brain-mental health associations in the UK Biobank
- A Genome-Wide Association Study Identifies Novel Genetic Variants Associated with Knee Pain in the UK Biobank (n = 439,743)
- Serum proteins predict incident inflammatory bowel disease 14 years before diagnosis: a machine-learning and Cox modelling in the UK Biobank
BibTeX
@misc{leonard2026gctag,
title = {{GCTAg: scalable mixed-model analysis for biobank-scale agricultural cohorts}},
author = {Alexander S. Leonard and Qiongyu He and Natasha Watson and Naveen Kumar Kadri and Hubert Pausch},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.03815},
eprint = {2610.03815},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.03815},
}
RIS
TY - UNPB
TI - GCTAg: scalable mixed-model analysis for biobank-scale agricultural cohorts
AU - Alexander S. Leonard
AU - Qiongyu He
AU - Natasha Watson
AU - Naveen Kumar Kadri
AU - Hubert Pausch
PY - 2026
DA - 2026/10/01
PB - arXiv
DO - 10.48550/arxiv.2610.03815
UR - https://arxiv.org/abs/2610.03815
AB - Genome-wide association studies identify genomic variants associated with traits. Mixed linear model association (MLMA) methods using a whole-genome relationship matrix, such as those implemented in GCTA, are powerful but computationally expensive. Here, we remove key memory and CPU bottlenecks in GCTA, reducing REML memory usage by nearly 75% and substantially accelerating MLMA by orders of magnitude in biobank-scale cohorts while preserving exactness. We further exploit relatedness in the mapping cohort through a reduced-rank Woodbury matrix approach, delivering further orders of magnitude performance gains with controlled genomic inflation. Native on-the-fly dominance recoding also eliminates slow I/O-operations on intermediate files, enabling efficient additive and dominance MLMA analyses in large agricultural cohorts.
ER -