(a) Basic demographic information on 119 individuals used to train the transcriptional signature models (‘training’) and the entire cohort of 251 participants (‘total’). Reported are the count (N) and percentage of non-missing values per each category. For each variable in parentheses, we reported the p-values for significant differences in variable distribution between the training group and the group not included in the model. (b) List of variables collected for the current study. DD: daily diary; SD: sleep diary; EAR: coded from Electronically Activated Recorder; YR: youth reported; PR: parent reported; CD: census data; GC: glucocorticoid; SD: standard deviation; α: Chronbach's alpha measuring reliability as the average correlation between scale items, as a function of the number of items included in the scale, N initial – sample size used for differential gene expression analysis and building transcriptional signatures (mean, SD, and α are reported for this subset), N longitudinal – number of samples with measurements from two time points available, N expanded – final sample size used for transcriptional signature and GxE validations. (c) Evaluation of transcriptional signatures derived using elastic net regression. For each variable, we report Pearson’s correlation coefficient, p-value, cross-validated percent variance explained, and sample size for the training dataset (columns 2–5), and Pearson’s correlation coefficient, p-value, and sample size for the entire cohort (columns 6–8). (d) Longitudinal replication of transcriptional signatures. For each variable, we report Spearman’s correlation coefficient between longitudinal change in observed variable and change in the transcriptional signature, p-value, coefficient of variation of longitudinal change in observed variable, and sample size. (e) Differentially expressed genes (DEGs) for all psychosocial variables (10% FDR; N: sample size). (f) Full DESeq results for differential gene expression analyses for psychosocial variables. (g) Correlations between z-scores of GxE interaction effects measured using transcriptional signatures and observed data. For each variable, we report Spearman’s correlation coefficient and its p-value, and odds ratio and p-value from Fisher’s test for enrichment of nominally significant results (permutation-corrected p-value<0.01) between GxE expression quantitative trait locus (eQTL) testing using transcriptional signatures and observed data. (h) Correlations between the top three principal components of covariate matrix and individual covariates (Pearson's product-moment correlations with numeric variables, polyserial correlations with bivariate variables; ns: correlation p-value>0.05). (i) DEGs associated with blood cell composition (10% FDR; N: sample size). (j) Full DESeq results for differential gene expression analyses for blood composition variables.