Fig. 3: Robustness of main exposome-omics associations.

Comparison of effect sizes of the 1170 exposome-omics associations of the main model, that includes all children (N = 1301) and is adjusted for child’s zBMI and ancestry, vs. effect sizes of alternative models. Each triangle represents an association. The x-axis represents the effect size of the exposure on the omics feature in the main model, while the y-axis represents the effect size in the alternative model. The percentage change in effect size between models is calculated as indicated in the Supplementary Information. A Alternative model included all covariates but was restricted to European ancestry children (N = 1171). No major differences were observed. B Alternative model included all children and was unadjusted for child’s zBMI. Exposure-omics associations with a percent change between models above 100% are coloured in red, and include proteins and child lipophilic chemical pollutants, as listed in the table. C Forest-plots showing the fixed- and random-effects inverse variance weighted meta-analyses of illustrative exposure–omics associations: maternal Cd levels and child DNA methylation at CpG cg19089201 (MYO1G gene) (n = 1173), with consistent effects across cohorts; child Cu levels and child CRP levels in plasma (n = 1170), with consistent effects across cohorts; maternal Mo levels and child DNA methylation at CpG cg08379738 (DENND1C gene) (n = 1173), driven by one of the cohorts (BiB); humidity in childhood (1 month before sampling) and child serum serotonin levels (n = 1198), driven by one of the cohorts (MoBA). Each cohort is represented by a point estimate, bounded by the 95% confidence interval (CI) for the effect and the cohort weight as a grey square. The 95% CI from the fixed and random effects meta-analysis are shown as diamonds. The effect size is reported as a log2 fold change (log2FC) of the omics, or difference in methylation levels, for interquartile range (IQR) of continuous exposure variables.