Figure 1. Association analyses of discovered rare functional variants in autoimmune diseases.

We define rare functional variants as MAF<0.5% in 17,019 controls and predicted non-synonymous, premature stop or splice site annotation. Quantile-quantile plots compare observed versus expected test-statistic distributions, with shading indicating 99% confidence intervals. Full results are available in Supplementary Data. Each of six individual diseases, and all autoimmune diseases combined, were tested as phenotypes.

a. Gene based C-alpha test (25 genes by 7 phenotypes, n=41,911 subjects) allowing for both risk and protective effects for rare functional variants (n = 41,911 subjects). Singleton variants pooled into a single binomial count per phenotype.

b. Gene based burden tests (25 genes by 7 phenotypes, n=41,911 subjects) comparing summed allele counts for rare functional variants in cases versus controls with Fisher’s exact test.

c. Conditional gene based burden test (25 genes by 6 phenotypes, n=32,806 subjects): rare functional variant allele counts are summed for each individual per gene and introduced in a logistic regression, including Immunochip covariates for multiple independent top (common) variant signals selected on the basis of a stepwise regression (down to P>10⁻⁴). The psoriasis phenotype was not tested as most samples do not have Immunochip data.

d. UNIQ-cases tests (25 genes by 7 phenotypes, n=41,911 subjects) that compares the number of rare functional variants only observed in cases with the distribution of this value upon random permutation (10,000 times) of the phenotypes.

e. UNIQ-controls, same as e but for rare functional variants uniquely observed in controls.