Background: Results from recent, high-powered DNA methylation (DNAm) studies have underscored the potential for EWAS to yield robust, credible findings that identify genomic regions associated with complex trait... [ view full abstract ]
Background: Results from recent, high-powered DNA methylation (DNAm) studies have underscored the potential for EWAS to yield robust, credible findings that identify genomic regions associated with complex trait pathophysiology or risk liability; however, issues of replicability have prompted researchers to identify practices that may be contributing to non-overlapping results. Covariates such as cell-type heterogeneity and substance use have received the most attention, but the impact of phenotype measurement and definition has been understudied. For example, EWAS of major depression (MD) have relied on depressive symptom (DSx) scores and MD case status to identify genomic regions associated with depression, but the question of whether these two broad depressive phenotypes both tap into the same latent construct is unresolved.
Objective: Evaluating the similarity of DNAm patterns associated with both DSx and MD case status using the same sample may provide evidence regarding the relationship between self-reported DSx and phenotype based on a clinical diagnosis.
Method: A co-twin control design was leveraged to assess genome-wide DNAm in N=75 monozygotic twin pairs (N=150 twins; ages 15-20 years; 73% female). Data from the Infinium HumanMethylation450 Beadchip was processed in the R environment using BioConductor packages. All samples passed rigorous quality control metrics, and poorly performing probes were identified and removed. Phenotypic data related to psychopathology, substance use, and demographic information were collected electronically using self-report questionnaires. DSx were measured using the Short Mood and Feelings Questionnaire, which is validated for use in both juvenile and adult populations. Early-onset MD case status was assessed using an extended version of the self-report Composite International Diagnostic Interview-Short Form (CIDI-SF). Linear regression methods were used to identify DNAm sites associated with an MD phenotype.
Results: One hundred thirty-six sites were differentially methylated by early-onset MD case status (1% False Discovery Rate). These sites mapped onto 84 genes, and enrichment analysis revealed a significant portion of these genes were associated with neuro-related functions, such as axon projection, neural differentiation, and synaptic signaling. No significant results were identified in the quantitative DSx analyses. Association sign tests revealed that only a minority of probes significant in the MD case analysis had the same direction of effect compared to DSx (4/136).
Conclusions: Significant differences in DNAm exist by early-onset MD case status. DSx and MD case status yielded exceptionally different results, which supports the notion that DSx and MD case status are disparate phenotypes and that poor replicability may be tied to this problem.
Psychopathology (e.g., Internalizing, Externalizing, Psychosis) , Statistical Methods , other
