The approximate numbers of cases and controls were:
Study
Platform
Cases
Controls
NIDDK (UK)
Illumina 550K
4,000
4,000
WTCCC (UK)
Affymetrix 500K
2,000
3,500
GAIN/NIMH (US)
Affymetrix 500K
1,800
1,800
Since the studies used two different platforms (with ~9% of the total number of SNPs in common), extensive use of imputation was required. Further details are available via the snpMatrix documentation.
SNPs were excluded if they failed one (or more) of the following QC criteria:
Minor allele frequency (case or control) < 0.01
Call rate (case or control) < 0.95 (T1DGC and WTCCC) or signal clouds insufficiently separated (GoKinD)
Chi-squared test for deviation from HWE (control) > 25
Version 18, 2009-05-11 14:47:12
Further Information
The paper acompanying this meta-analysis can be found in Nature Genetics 2009 - Genome-wide association study and meta-analysis find that over 40 loci affect risk of type 1 diabetes; Barrett et al. (PMID 19430480).
A curated list of T1D susceptibility loci is available for browsing. Support information, such as gene lists and genomic sequences for all human loci can be downloaded from the Downloads section of T1DBase.
Version 2, 2009-05-28 14:20:43
Top SNPs
The top SNPs in the study were determined by: selecting those with a stratified 1df p-value > 1e-03; removing any in the MHC region (here defined as chr6 25.7Mb-33.3Mb); removing any which failed QC (cf criteria in previous section). Of the resulting set of SNPs, if there was more than one SNP within a 500kb region, the one with the lowest p-value was selected.
For the SNPs that remained after this filtering process, LD regions were calculated with the HapMap CEU Founders dataset, in snpMatrix, using different D' and r² thresholds. Unless otherwise noted, LD was calculated using the following criteria: minor allele frequency > 0.05; call rate > 0.9; deviation from HWE < 5.
If a gene is wholly or partially within a region at a given threshold, it is displayed in the 'Genes' column; for these calculations the gene boundaries have been extended by 3kb in either direction, to encompass potential transcriptional effects. Of the SNPs that were discarded because they were within 500kb of a SNP with a lower p-value, if any are in LD with the SNP with the lowest p-value, at a given threshold, they are listed in the 'SNPs' column.
The table below is available for download as a text file.