Brass

Case-control collections are listed in Table 1 and described in detail in our previous studies^{6 (link),7 (link),11 (link),13 (link),19 (link)}. Collections were composed entirely of individuals of selfdescribed European ancestry, and all cases either met the 1987 American College of Rheumatology (ACR) criteria for diagnosis of RA^{50 (link)} or were diagnosed by board-certified rheumatologists. All RA cases were further limited to anti-CCP+ patients, or RF+ patients if anti-CCP status data were missing. The BRASS RA samples have been used in our previous 100K GWAS^{6 (link)}, but here are presented for the first time genotyped with the Affymetrix 6.0 array. In the current study, controls were matched to BRASS RA cases using principal components analysis from GWAS data from three separate studies: controls from a multiple sclerosis GWAS^{31 (link)}, controls from an age-related macular degeneration GWAS⁵¹ , and controls from a myocardial infarction GWAS^{52 (link)}. WTCCC collection controls included the 1958 birth and National Blood Service cohorts as well as non-autoimmune disease cases (bipolar disorder, cardiovascular disease, hypertension and type 2 diabetes)^{15 (link)}. All GWAS collections except WTCCC were restricted to control subjects matching cases using principal components analysis of GWAS data. All replication sample collections were composed of epidemiologically and/or geographically matched cases and controls, except NARAC II collection, which was case-control matched based on genotypes at a set of ancestry-informative markers as previously described^{11 (link)}. The eight replication samples included: (1) CCP positive cases and controls from Halifax and Toronto (CANADA-II)^{13 (link)}; (2) RF positive Dutch cases from Groningen and Nijmegen, together with geographically matched controls^{53 (link)}; (3) CCP positive Dutch cases and controls collected from the greater Amsterdam region (GENRA)^{54 (link)}; (4) North American RF positive cases and controls matched on gender, age, and grandparental country of origin from the Genomics Collaborative Initiative (GCI)^{4 (link)}; (5) CCP or RF positive Dutch cases and controls from Leiden University Medical Center (LUMC)^{5 (link)}; (6) CCP positive cases drawn from North American clinics and controls from the New York Cancer Project (together this collection is called NARAC-II)^{7 (link)}; (7) CCP or RF positive cases recruited at multiple sites in the United Kingdom by the United Kingdom Rheumatoid Arthritis Genetics (UKRAG) collaboration^{9 (link)}; and (8) CCP or RF positive cases identified by chart review from the Nurses Health Study (NHS) and matched controls based on age, gender, menopausal status, and hormone use^{55 (link)}. We used available SNP data from this and previous studies to identify genetically identical samples from the same country^{13 (link)}; we assumed these represented duplicated individuals and removed them. Institutional review boards at each collection site approved the study, and all individuals gave their informed consent.

The BRASS GWAS collection was genotyped on the Affymetrix Genechip 6.0 platform at the Broad Institute (Boston, USA). All other GWAS collections were genotyped as previously described^{7 (link),13 (link),19 (link)}. Genotype data for GWAS samples from RA and other disease studies were obtained with permission from the investigators and/or disease consortia. Additional shared control GWAS genotype data were obtained from the NIMH (USA) through a formal application and approval process (part of the BRASS collection), and from the Illumina iControls database (NARAC III). For each GWAS collection, quality control (QC) was implemented in cases and in each control cohort separately, and then again in the merged collection data. QC steps included filtering SNPs and individuals with >5% missing data, followed by filtering SNPs with minor allele frequency (MAF) <1% and a Chi-squared test of Hardy Weinberg equilibrium P_HWE<10⁻⁶. For the WTCCC collection, genotyped on the older Affymetrix 500K platform, we implemented more stringent QC (>1% missing data, MAF<1%, P_HWE<10⁻⁵). We then used individual-pairwise identity-by-state estimates to remove occasional related and potentially contaminated samples. Data processing and QC were performed in PLINK^{56 (link)}. Additional details are described in the Supplementary Note online.
The 34 SNPs chosen for replication, as well as proxy SNPs, were directly genotyped in each of eight collections (Table 1 and Supplementary Note online). Canada samples were genotyped on the Sequenom iPlex platform at University of Toronto, Mount Sinai Hospital and University Health Network (Toronto, Canada); Dutch and GENRA samples were genotyped on the Sequenom iPlex platform at the Broad Institute; UKRAG samples were genotyped on the Sequenom iPlex platform at The University of Manchester, UK; GCI and LUMC samples were genotyped by kinetic PCR at Celera (Alameda California, USA); NARAC II samples were genotyped on the Sequenom iPlex platform at the NIH (USA); and NHS samples were genotyped on the Biotrove OpenArray platform at the Channing Laboratory, Harvard Medical School (Boston, USA). QC filter criteria for SNPs in each replication or validation collection were 10% missing data, 1% MAF and P_HWE<10⁻³. If a given SNP failed in genotyping or QC in a collection, a proxy SNP (r²>0.8) with the least missing data (if available) was used. See Supplementary Note online for details.