Fig. 6

EpiDISH improves sensitivity of a smoking EWAS in blood. a Distribution of cellular proportions (weight) for the main blood cell subtypes in the 152 whole blood samples from Teschendorff et al., as inferred using EpiDISH. b Heatmap of P-value associations between significant principal components (PC) and various factors, including Smoking Status, Bisulfite Conversion Efficiency (BSC), sentrix ID and position, and cellular proportion for 6 blood cell subtypes (eosinophils were not considered due to weights being effectively all zero). c Quantile-quantile plot of all 450k probes passing quality control from a supervised analysis against smoking-pack-years only adjusted for sentrix ID (“No adjustment”). The number of CpGs passing an FDR < 0.05 threshold are given, and defined to be smoking-associated differentially methylated CpGs (sDMCs). d Quantile-quantile plot of all 450k probes passing quality control from a supervised analysis against smoking-pack-years adjusted for sentrix ID and blood cell subtype proportions as estimated using EpiDISH (“Adjusted (EpiDISH)”). The number of CpGs passing an FDR < 0.05 threshold are given, and defined to be smoking-associated differentially methylated CpGs (sDMCs). e Among the 34 and 70 sDMCs identified in c and d, respectively, we indicate the numbers of these that map to a selected set of 5 well-known and validated smoking-associated genes. This subset derives from a set of 15 gold-standard smoking-associated genes, as curated by a review of the literature by Gao et al. [30]