gpu-coloc
Summary¶
gpu-coloc is a GPU-accelerated bioinformatics tool designed for genetic colocalization analysis. It enables high-throughput colocalization across millions of traits, making it feasible to colocalize large-scale GWAS datasets with comprehensive molecular QTL datasets (including eQTL, sQTL, and pQTL).
Overview¶
gpu-coloc was developed to address the computational bottleneck of standard colocalization methods when analyzing hundreds of metabolic or molecular traits against thousands of GWAS traits. By leveraging GPU acceleration, it significantly reduces the time required to perform pairwise colocalization analyses. In the study by Tambets et al. (2026), gpu-coloc was used to perform colocalization of 86,886 signals from a multi-ancestry metabolic trait GWAS meta-analysis against summary statistics for up to 7,228 traits from various biobanks and QTL databases, identifying 932,864 colocalization events (using PP.H4 > 0.9).
Applications¶
- Large-scale Colocalization: Utilized to perform systematic phenome-wide colocalization of metabolic trait GWAS signals against molecular QTL and complex disease traits.
- eQTL and pQTL Integration: Integrated with the eQTL Catalogue, INTERVAL BioResource, and UKBB Pharma Proteomics Project (UKB-PPP) to identify target genes and molecular mechanisms.
Related References and Contributors¶
- Colocalization Posterior Probability (CLPP) — a posterior metric used to prioritize shared causal signals.
- Alasoo Lab — contributor in functional genomics and statistical colocalization.
- Palta Lab — contributor in population-scale genetic analysis.
Citations¶
- Jesse, M., Riet, A.-E. & Alasoo, K. (2025). Ultra-fast genetic colocalisation across millions of traits. bioRxiv preprint. DOI: 10.1101/2025.08.25.672103
- Tambets, R. et al. (2026). Genetic analysis of circulating metabolic traits in 619,372 individuals. Nature. DOI: 10.1038/s41586-026-10532-5. Source paper: s41586-026-10532-5.pdf