Fig. 5: Cigarette smoking confounds associations between psychosocial risk factors and DNA methylation.

a, b show the scatterplot of the –Log₁₀(p) values from an epigenome-wide association studies (EWAS) of adverse childhood experiences (ACEs; x-axis) plotted against –Log₁₀(p) from an EWAS of ACEs controlling for pack-years smoked at age-38 in the Dunedin Study (a) and age-18 in the E-Risk Study (b). c, d show the same scatterplot but substituting SmPEGS at age-38 in Dunedin (c) and age-18 in E-Risk (d) for pack-years smoked. Dashed lines represent the genome-wide significant cutoff levels. Blue points represent probes significant in the EWAS of ACEs. Had EWAS hits for ACEs remained significant after controlling for smoking (a, b) or for Smoking methylation Polygenic Score (SmPEGS) (c, d), the blue points would appear in the upper right-hand quadrant. In both cohorts, controlling for smoking or using the SmPEGS attenuates the association between ACEs and DNA methylation to non-significance; hence the blue points appear in the lower right-hand quadrant