Test. More specifically, we tested for over-representation of genome-wide significant SNPsTest. More specifically, we tested

Test. More specifically, we tested for over-representation of genome-wide significant SNPs
Test. More specifically, we tested for over-representation of genome-wide significant SNPs (i.e. association of og10 p-value > 7.3) within specific enhancers. We adjusted for multiple testing using a Bonferroni-correction accounting for the number of annotations tested.First, we overlaid genome coordinates of enhancers in PC cell lines, as defined previously, with genomic coordinates of all SNPs in the PC iCOGS dataset [29]. To visualize differential Nutlin-3a chiral chemical information enrichment patterns of specific epigenetic markers with respect to their genetic association with PC risk we generated stratified Q-Q plots which is a method for visualizing the enrichment of statisticalZuber et al. BMC Genomics (2017) 18:Page 4 ofTable 1 SNPs mapping to epigenetic marks used to define enhancers in prostate and breast cancerCell line/Tumor type LNCaP Epigenetic marks H3K27Ac Key transcription factora —- AR LNCaP H3K27Ac + MED12 —- AR VCaP H3K27Ac + BRD4 —- AR VCaP BRD4 —- AR VCaP PC H3K27Ac H3K27Ac + MED12 + BRD4 —- H3K27Ac + BRD4 —- —- AR PC MCF7 AR —- ER MCF7 H3K27Ac —- ER MCF7 BRD4 —- —- ER MCFaNumber of SNPs within the intervals covered by the array: iCOGS 1605 669 685 279 587 342 859 23 3896 82 49 496 8783 4296 19270 6710 280 34 158 PRACTICAL 3092 1274 1271 541 983 502 1595 60 8440 130 46 1403 21058 10607 48028 16206 495 66 371 BCAC 13482 5664 5428 2310 4218 2148 7984 233 38150 618 248 5950 93969 45937 215997 69743 2617 282 1638 All 13503 5671 5442 2316 4239 2154 8066 233 38444 619 249 5967 94858 46222 217382 70058 2641 282ERBinding information for key transcription factors such as androgen receptor (AR) or estrogen receptor (ER) where used alone or in combination with the epigenetic mark profiles in order to assess their capacity to refine enrichment of risk SNPsassociation relative to that expected under the global null hypothesis [18]. Q-Q plots show that SNPs within regions with different genomic features (H3K27ac, BRD4, and MED12, or a combination of these) had different enrichment patterns compared to all SNPs (Fig. 2a). The SNPs contained in common PC enhancers, and therefore characterized by BRD4 and MED12 binding, and a long stretch of H3K27Ac had PubMed ID:https://www.ncbi.nlm.nih.gov/pubmed/25768400 lower pvalues than SNPs contained in enhancers identified in VCaP cells by mapping long stretches of H3K27Ac and BRD4 binding. SNPs associated with PC risk were more enriched within BRD4 binding sites alone than within H3K27Ac sites or H3K27Ac/MED12 overlapping sites in LNCaP. In addition, we focused on SNPs achieving genome wide significance and compared overrepresentation of these SNPs mapping to the above-described enhancers (Additional file 22: Table S1). 12 and 3 of the SNPs contained in PC enhancers achieved genome-wide significance in the iCOGS and in the PRACTICAL GWAS respectively. SNPs that achieved significance in iCOGs are listed in Additional file 22: Table S2. These results highlight that combining generic epigenetic marks such as H3K27Ac with generic epigenetic readers such as BRD4 and with MED binding increases the capacity of capturing SNPs associated with PC.Importantly, to rule out possible confounding factors, we first randomly pruned the SNPs, selecting one representative SNP per LD block. The random pruning did not change the enrichments patterns caused by the functional annotations (Additional file 22: Figure S1). Secondly, in order to rule out that the enrichment merely results from the non-independence of the SNPs in the enhancer regions or other confounding features.