Fig. 2

Characterization of autism candidate gene sets. a We first classified the 11 gene sets used in the study into three groups: no effects (light grey), weak effects (pale orange), and strong effects (light pink). We see little overlap in the individual gene sets themselves (middle panel). The total number of genes in each set also varies (right panel), and is negatively correlated with the average effect size (r_s = −0.69). The bars represent the total number of genes within each gene set, and the black dots the average disease effect size. b Control gene sets overlap significantly with missense genes (333 genes, hypergeometric test p = 2.54e-6), common SNP gene sets (11 genes, p = 4.5e-3), and the loss-of-function (LoF) SNV gene sets (71 genes, p = 3.2e-3), but not the CNV gene sets (14 genes, p = 0.03). Missense and common SNPs overlap significantly (four genes, p = 2.4e-4). However, loss-of-function SNVs do not overlap significantly with either common (four genes, p = 0.62), missense (68 genes, p = 0.75), or CNVs (three genes, p = 0.37). c Common biases that affect studies are gene length and number of functional annotations. The average standardized rank (± standard error) of genes with respect to these properties shows that the “rare” disease sets contain longer genes but are not much more multifunctional than random