Sex differences in disease risk from reported genome-wide association study findings

Abstract

Men and women differ in susceptibility to many diseases and in responses to treatment. Recent advances in genome-wide association studies (GWAS) provide a wealth of data for associating genetic profiles with disease risk; however, in general, these data have not been systematically probed for sex differences in gene-disease associations. Incorporating sex into the analysis of GWAS results can elucidate new relationships between single nucleotide polymorphisms (SNPs) and human disease. In this study, we performed a sex-differentiated analysis on significant SNPs from GWAS data of the seven common diseases studied by the Wellcome Trust Case Control Consortium. We employed and compared three methods: logistic regression, Woolf's test of heterogeneity, and a novel statistical metric that we developed called permutation method to assess sex effects (PMASE). After correction for false discovery, PMASE finds SNPs that are significantly associated with disease in only one sex. These sexually dimorphic SNP-disease associations occur in Coronary Artery Disease and Crohn's Disease. GWAS analyses that fail to consider sex-specific effects may miss discovering sexual dimorphism in SNP-disease associations that give new insights into differences in disease mechanism between men and women.

Fig. 1

The analysis pipeline for each of seven diseases. We performed sequential filtering steps and applied stringent quality control criteria at both the SNP and individual level. We then performed disease-control association analysis on each sex separately

Fig. 2

Chi-square statistics for T1D SNP rs480092 from 10,000 random label permutations. Lines are drawn to indicate the actual male (blue) and female (red) Chi-square statistics from the disease genotype data. The sex difference p value is 0.0191, and is calculated from the proportion of random permutations where the absolute value of the difference between male and female Chi-square statistics is more extreme than the actual difference we observed from the real data (upper left corner and lower right corner of the plot, shaded in green). The position corresponding to the ordered pair of the actual male and female Chi-square statistics is highlighted in yellow (color figure online)

Fig. 3

Genome-wide scan for seven diseases highlighting sex-specific SNPs. The log10 of the disease association p values for SNPs are plotted against chromosomal position for each of seven diseases. The separate male p value and female p value are plotted for each SNP. The 280 SNPs considered for sex-specific effects are colored in green. SNPs with a sex-difference p value <0.05 are colored in red and blue, with the female p value in red and the male p value in blue. Red open upper arrows and dark blue lower arrows highlight these SNPs as well, respectively. All panels are truncated at 1 × 10⁻¹⁵, although some markers exceed this significance threshold (e.g., chromosome 6 in T1D and RA). The red and blue horizontal lines indicate the Bonferroni correction for each disease (color figure online)