A GWAS Summary Statistics Library for Cross-Trait Analyses

Abstract

Genome-wide association studies (GWAS) have generated vast amounts of genetic data, yet inconsistent formats and fragmented metadata continue to limit large-scale reuse. This project directly addresses that gap by contributing to the foundation of a standardized GWAS summary statistics library, optimized for high-performance computing (HPC) environments. The overarching aim is to enable systematic, scalable access to GWAS data, allowing researchers to perform a range of common analyses without manual intervention. More sophisticated methods can still be applied separately, but this work ensures the data is gathered and formatted in a consistent, usable way. To achieve this, an automated harmonization pipeline was developed to resolve formatting and structural differences in publicly available GWAS files. An initial audit revealed substantial heterogeneity in file formats, with only 42% of datasets following the most common structure; the remainder required extensive correction. Post-harmonization, all files were standardized with 100% consistency, supporting automation of downstream analyses. Benchmarking of storage requirements and query speed across various summary statistic file formats showed that VCF-based formats—especially compressed and indexed VCF—offered the best performance for common queries, while uncompressed VCF outperformed others for large, multi-variant extractions. These findings inform practical format choices for efficient data integration at scale. A pilot colocalisation analysis targeting the ALDH2 locus confirmed the infrastructure’s utility, revealing strong evidence of pleiotropy across skin diseases. Together, these results demonstrate a scalable framework for the systematic collation and retrieval of GWAS summary statistic data and their use in cross-trait genetic discovery.