Table 1.
Description of the GHS study population.
Table 2.
Number of gene expression-by-SNP associations at various levels of significance.
Figure 1.
Number of eQTLs according to the significance threshold adopted and corresponding cis/trans eQTL ratio.
The vertical arrow indicates the study-wise level of significance correcting for the number of hypotheses tested. Some eQTLs being associated with both cis and trans-acting eSNPs, the sum of cis and trans eQTLs is greater than the total number of eQTLs.
Figure 2.
Comparison of the distributions of P-values of cis eQTLs reported as significant in three previous association studies with P-values observed in GHS for the same eQTLs.
For each of the 3 comparisons, we selected in GHS the subset of gene expressions claimed as significant in the study of comparison. Only autosomal genes were considered in these comparisons. The data used to generate this figure are provided in Files S2–S4. See also footnote of Table S3 for details.
Table 3.
Number of cis eQTLs identified in previous studies and replicated in GHS.
Figure 3.
Comparison of the heritability of cis eQTLs estimated in the SAFHS study with the R2 of the corresponding cis eQTLs in GHS.
Data were extracted from Supplementary Table 4 in Göring et al. [4] and comparisons were restricted to genes having a corresponding gene symbol in GHS. Heritability in the SAFHS was estimated by linkage analysis and accounts for the whole variability at a locus while R2 refers to a single eSNP (the best eSNP) and therefore underestimates the global variability affecting gene expression at a locus. The data used to generate this figure are provided in File S5. The median R2 was globally lower than the heritability, consistent with the fact that the R2 is referring to a single SNP whereas heritability reflects the whole genetic variation at a locus.
Table 4.
Loci identified in GWAS of circulating lipids – associations of lead/tag SNPs with phenotypes and expression, and of expression with phenotype in GHS.
Figure 4.
The loci affecting CDKN2B expression and CAD on chromosome 9p21 are independent.
The lead SNP rs1333049 generally reported at the CAD locus was not present on the Affymetrix 6.0 array, we therefore selected its best proxy, rs10757272 (position 22078260, r2 = 0.9 with rs1333049), using SNAP (https://www.broadinstitute.org/mpg/snap). Positions of genotyped SNPs are shown using a green link and position of the proxy SNP, rs10757272, is represented by a green triangle. The red curve reflect the –log10(P-value) for the association between SNPs and CDKN2B expression. The LD (r2) between pairs of SNPs is shown at the bottom of the figure using a range of colors between white (r2 = 0) and black (r2 = 1). The CDKN2B and CAD-associated SNPs are located in different blocks of LD strongly suggesting that the genetic effects on CDKN2B expression and CAD are independent.
Table 5.
Number of expression traits associated with risk factors and cis eSNPs.
Figure 5.
Effect of the best cis eSNP and smoking on expression of smoking-related eQTLs.
The proportion of variability of expression explained by the best cis eSNP varied from 3.1% for CLEC10A to 27.2% for GFRA2 while the proportion explained by smoking varied from 2.8% for SMAD6 to 21.6% for SASH1. The lowest P-value for interaction between SNP and smoking was 0.02 for STAB1.
Table 6.
Subsets of expression traits showing robust independent association with the different risk factors in the validation sample.
Table 7.
Smoking-related expression traits: association of expression with cis-acting eSNP, smoking and extent of atherosclerosis.