Fig. 3: Mediation proportions (\(\widehat{{{{{{{{\rm{MP}}}}}}}}}\)s) for transcripts in cis mediating DNAm-to-trait effects.
a Flowchart describing the selection of DNAm-trait pairs retained for mediation analyses. Among the total of 2,480,677 pairs tested, 2069 pairs (orange) with a significant total causal effect (PT) and at least 1 causally-associated transcript were assessed for mediation (b, c). 554 pairs without any transcript causally linked to the DNAm site (green) were included in the calculation of the overall mediation. Pairs without any transcript in the cis region (pink) were omitted in mediation analyses as were pairs without sufficient instrumental variables (IVs, purple). b\(\widehat{{{{{{{{\rm{MP}}}}}}}}}\)s by trait where error bars denote the 95% CI, and the grey vertical bar shows the mean \(\widehat{{{{{{{{\rm{MP}}}}}}}}}\) across the traits (\(\widehat{{{{{{{{\rm{MP}}}}}}}}}\)s per trait were derived by regressing \({\hat{\theta }}_{D}\) against \({\hat{\theta }}_{T}\)). Only traits with ≥10 DNAm-trait pairs with detectable mediation are displayed (41 traits with number of evaluated pairs indicated in parentheses), colour-coded by their physiological category as defined in the legends of c and d. c Detectable mediation: All DNAm-trait pairs assessed in the mediation analyses (2069) with traits being grouped into 10 physiological categories. The global \(\widehat{{{{{{{{\rm{MP}}}}}}}}}\) in percentage with 95% CI is shown in the plotting area and individual category \(\widehat{{{{{{{{\rm{MP}}}}}}}}}\)s in the legend. d Overall mediation: Same analysis as in c, but including all 2623 DNAm-trait pairs with at least 1 transcript present in the cis region. For these additional pairs, the direct effect was set to the total effect.
