Fig 1.
The genomic locations of the 147 prostate cancer risk SNPs.
Fig 2.
Various samples and genomic data used in this paper.
Fig 3.
The results of eQTL mapping for the 147 prostate cancer risk SNPs in the Mayo (histologically normal), TCGA (tumor) and FH (tumor) data.
A, the quantile-quantile plot of SNP-gene association p-values. B, the Venn diagram of the eQTLs and eGenes identified in the Mayo histologically normal set and the eQTLs and eGenes confirmed in the TCGA and FH tumor set. C, the scatter plot of eQTL effect sizes in Mayo benign samples and TCGA tumor samples. D, Standardized gene expressions grouped by genotypes for three examples of eQTL-eGene. The top panel is rs12653946-IRX4 identified in both tumor and tumor-adjacent histologically normal samples, the middle panel is rs4430796-HNF1B only identified in adjacent normal samples, and bottom panel is rs11135910-EBF2 identified in tumors only.
Table 1.
Summary of cis-eQTL mapping results for 147 prostate cancer risk SNPs in three datasets.
Fig 4.
The results of meQTL mapping for the 147 prostate cancer risk SNPs in the FH (tumors) and TCGA (tumors).
A, the quantile-quantile plot of SNP-CpG association p-values. B, the Venn diagram of the meQTL identified in the FH tumor set and the meQTL confirmed in the TCGA tumor set. C, beta values of CpG methylation grouped by genotypes for an example meQTL-CpG association.
Table 2.
Summary of cis-meQTL mapping results for 147 prostate cancer risk SNPs in two datasets.
Fig 5.
Genetic regulation of gene expression overlaps with genetic regulation of CpG methylation in the 147 PrCa risk SNPs.
A, the Venn diagram of eQTL and meQTL identified in the FH tumor set. B, the histogram of p-values when applying the two-stage least squares method to FH SNP-CpG-gene expression triplets when SNPs are both eQTL and meQTL. C, the scatter plot of genetic associations with gene expression and genetic associations with CpG methylation for SNPs identified as eQTL and meQTL in the FH tumor set and CpGs located in the same gene of gene expression. D, the Venn diagram of eQTL and meQTL identified in the TCGA tumor set. E, the histogram of p-values when applying the two-stage least squares method to FH SNP-CpG-gene expression triplets when SNPs are both eQTL and meQTL. F, the scatter plot of genetic associations with gene expression and genetic associations with CpG methylation for SNPs identified as eQTL and meQTL in the TCGA tumor set and CpGs located in the same gene of gene expression.
Fig 6.
Altered DNA methylation in genes which lose genetic control in tumors when compared to benign samples.
A, Number of differentially methylated probes (DMP) in the eGenes in benign samples only and eGenes in both tumor and benign samples. B, Number of differentially methylated blocks in eGenes in benign samples only and eGenes in both tumor and benign samples. C, the percentage of genes associated with at least one meQTL-CpG pair within the genes.
Fig 7.
Three possible SNP-Methylation-gene expression relationships and representative results of mediation analyses from TCGA and FH.
Table 3.
Examples of the triplets (SNP, CpGs, eGene) with a mediation relationship (either SME or SEM) that is consistent in FH and TCGA datasets.
Fig 8.
An example of the mediation relationship: rs12653946 and IRX4 in chromosome 5.
A, residuals of gene expression~CpG versus genotype. B, gene expression versus residuals of CpG~genotype. C, residuals of CpG~gene expression versus genotypes. D, the genomic annotation maps by UCSD genome browser for the region with rs12653946 and IRX4 in chromosome 5.
Fig 9.
Another example of the mediation relationship: rs3096702 and NOTCH4 in chromosome 6.
A, residuals of gene expression~CpG versus genotypes. B, gene expression versus residuals of CpG~genotype. C, residuals of CpG~gene expression versus genotypes. D, the genomic annotation maps by UCSD genome browser for the region with rs3096702 and NOTCH4 in chromosome 6.