Fig. 2: Research questions and methodological workflow.

a The EBICglasso graph-theoretical model was used to identify bridging symptoms in the depression-anxiety inter-symptom network, and the factor analysis was further conducted to identify the common factor to characterize a general structure of these identified bridging symptoms, which was conceptualized as cb factor. b To probe whether this conceptualized cb factor had neural substrates, the edge-centric brain connectome-based feature (i.e., edge-centric functional connectivity, eFC) was calculated as a neural feature for training the eFC connectome-based predictive model (eCPM) to predict the cb factor scores. Here, this model was trained in the discovery sample, and the model performance was evaluated by validating in an independent validation sample and generalization samples. c Once revealing the predictive roles of eFC features, the inter-subject representation similarity analysis (RSA) was further conducted by correlating the eFC pattern (i.e., eFC connectomes at a given eFC) to the behavioral feature (i.e., cb factor scores), to probe how these eFCs characterized the cb factor. d By using the ACE model in an independent twin cohort, these eFCs that identified significant RS to the cb factor, were examined for heritability. Once the heritability was confirmed, the Allen Human Brain Atlas (AHBA) was used to test whether such RS could be predicted by regional gene expression patterns in the partial least square regression (PLS). Finally, if the connectome-transcriptional correlates (i.e., gene expression patterns) were found, these specific gene patterns were annotated by using multiscale normative biological atlases.