Fig. 4: Four significant CpGs identified by epigenome-wide meta-analysis and their associations with timing, type, and duration of maltreatment exposure in the Toddler Social Cognition and Adolescent Brain Imaging cohorts.

Toddler Social Cognition and Adolescent Brain Imaging: (A) and (C) Structure of maltreatment history for each individual. Five-step gradation (pink blocks) indicates the number of CM types each individual was exposed to at each age. Gray blocks indicate ages not yet reached by the individual. PA: physical abuse; EA: emotional abuse; SA: sexual abuse; NG: neglect; mother: biological mother; father: biological father; grandmo: grandmother; grandfa: grandfather; adpmo: adoptive mother; adpfa: adoptive father. Toddler Social Cognition and Adolescent Brain Imaging: (B) and (D) Sensitive-period analyses assessing maximal sensitivity by age of exposure (independent of maltreatment type/number) and by maltreatment type/number (independent of age of exposure). Random forest regression with conditional trees was used to evaluate the relative importance of maltreatment exposures during specific age windows (0–18 years) or specific maltreatment types/numbers for methylation at each of the four CpG sites. Relative importance was indicated by the degradation in model fit (increase in mean squared error, MSE) when each factor was permuted. *P < 0.05. Adolescent Brain Imaging: (E) Correlation plots illustrating significant associations between duration of maltreatment exposure and methylation levels at cg23172545_SERPINB9P1 and cg04794662_FOXP1.