Fig 1.
A. Initially, SNP association p-values are produced by an association test. Based on these values, top 1000 SNPs are selected and annotated to nearby genes. B. INRICH software is then used to detect over-represented gene-sets (empty circles denote all genes that were not detected by GWAS). Genes from top 20 such sets are retrieved, and only the ones overlapping with GWAS hits are analysed further. C. Using publicly available expression data from NP samples, this gene list is filtered to retain only differentially expressed genes. D. For each of these remaining targets, known eQTLs are assigned into bins based on effect direction (up- or down- regulating) and frequency distribution in our genotyping data (higher frequency in cases or in controls). Fisher's exact test is used to evaluate the observed distribution. 500 genes with equal or higher count of eQTLs are analysed in the same way, and statistic values generated from this control set are then compared with the target gene statistic to estimate the empirical significance.
Table 1.
Descriptive statistics of the participants who passed quality control.
Fig 2.
Manhattan plot from the DFAM analysis.
Fig 3.
Manhattan plot from the EMMAX analysis.
Table 2.
Top 30 associated SNPs from DFAM analysis.
Table 3.
Top 30 associated SNPs from EMMAX analysis.
Table 4.
Overlap between genes from INRICH analysis and expression data (from Plager et al. [32]).
Table 5.
Results from the eQTL analysis, Blood dataset.
Table 6.
Results from the eQTL analysis, MuTHER dataset.