We downloaded phenotype-associated SNPs and phenotype information from NHGRI GWAS catalogue database26 (link) on January 31, 2013. We selected significantly associated 4,676 SNPs (P < 5.0×10−8) corresponding to 311 phenotypes (other than RA). We manually curated the phenotypes by combining the same but differently named phenotypes into the single phenotype (eg. from “Urate levels”, “Uric acid levels”, and “Renal function-related traits (urea)” into “Urate levels”), or splitting the merged phenotypes into the sub-categorical phenotypes (eg. from “White blood cell types” into “Neutrophil counts” , “Lymphocyte counts” , “Monocyte counts” , “Eosinophil counts” or “Basophil counts”). Lists of curated phenotypes and SNPs are available at http://plaza.umin.ac.jp/~yokada/datasource/software.htm .
For each of the selected NHGRI SNPs and the RA risk SNPs identified by our study (located outside of the MHC region), we defined the genetic region based on ±25 kbp of the SNP or the neighboring SNP positions in moderate LD with it in Europeans or Asians (r2 > 0.50). If multiple different SNPs with overlapping regions were registered for the same NHGRI phenotype, they were merged into the single region. We defined “region-based pleiotropy” if two phenotype-associated SNPs shared part of their genetic regions or shared any UCSC ref gene(s) (hg19) partly overlapping with each of the regions (Extended Data Fig. 4a ). We defined “allele-based pleiotropy” if two phenotype-associated SNPs were in LD in Europeans or Asians (r2 > 0.80). We defined the direction of effect as “concordant” with RA risk if the RA risk allele also leads to increased risk of the NHGRI disease or increased dosage of the quantitative trait; similarly, we defined relationships as “discordant” if the RA risk allele is associated with decreased risk of the NHGRI disease phenotype (or if the RA risk allele leads to decreased dosage of the quantitative trait).
We evaluated statistical significance of region-based pleiotropy of the registered phenotypes with RA by a permutation procedure with ×107 iterations. When one phenotype had n loci of which m loci were in region-based pleiotropy with RA, we obtained a null distribution of m by randomly selecting n SNPs from obtained NHGRI GWAS catalogue data and calculating number of the observed region-based pleiotropy with RA for each of the iteration steps. For null distribution estimation, we did not include the SNPs associated with several autoimmune diseases which were previously reported to share pleiotropic associations with RA (Crohn's disease, type 1 diabetes, multiple sclerosis, celiac disease, systemic lupus erythematosus, ulcerative colitis, and psoriasis)2 (link).
For each of the selected NHGRI SNPs and the RA risk SNPs identified by our study (located outside of the MHC region), we defined the genetic region based on ±25 kbp of the SNP or the neighboring SNP positions in moderate LD with it in Europeans or Asians (r2 > 0.50). If multiple different SNPs with overlapping regions were registered for the same NHGRI phenotype, they were merged into the single region. We defined “region-based pleiotropy” if two phenotype-associated SNPs shared part of their genetic regions or shared any UCSC ref gene(s) (hg19) partly overlapping with each of the regions (
We evaluated statistical significance of region-based pleiotropy of the registered phenotypes with RA by a permutation procedure with ×107 iterations. When one phenotype had n loci of which m loci were in region-based pleiotropy with RA, we obtained a null distribution of m by randomly selecting n SNPs from obtained NHGRI GWAS catalogue data and calculating number of the observed region-based pleiotropy with RA for each of the iteration steps. For null distribution estimation, we did not include the SNPs associated with several autoimmune diseases which were previously reported to share pleiotropic associations with RA (Crohn's disease, type 1 diabetes, multiple sclerosis, celiac disease, systemic lupus erythematosus, ulcerative colitis, and psoriasis)2 (link).
