Unveiling latent neural mechanisms of depressive symptoms: Resting-state EEG biomarkers from symptom network and...

This study aimed to elucidate the neurophysiological mechanisms underlying depressive symptoms in university students by integrating symptom network analysis, resting-state EEG (rsEEG) source localization, and multimodal predictive modeling. We sought to identify EEG biomarkers driving symptom propagation and develop an interpretable neurocircuit-level framework.

A cohort of 489 university students (depression subgroup: n = 170 with BDI-II ≥ 14) was recruited. Symptom networks were modeled using a Mixed Graphical Model (MGM) to quantify cross-frequency EEG interactions (delta/theta/alpha/beta). Weighted minimum norm estimation (wMNE) localized the neural generators of aberrant oscillations. Key biomarkers from symptom network centrality and source-localized features were integrated via LASSO regression to construct a depression classification model, with performance evaluated by the area under the ROC curve (AUC).

1) Symptom network analysis: The depression subgroup exhibited elevated betweenness centrality in prefrontal delta bands (F7δ/F4δ: 2.38/1.768) and increased strength centrality in temporal delta (T3δ: 1.249), suggesting default mode network dysregulation. Central-parietal theta integration capacity was reduced (betweenness: -0.977 vs. 1.373 in controls). 2) Source localization: Aberrant oscillations localized to the prefrontal cortex, postcentral gyrus (BA 3-1-2), and inferior parietal lobule (BA40). Central theta/alpha abnormalities showed differential classification utility (Cliff's δ = 0.125-0.184). 3) Predictive model: LASSO-selected cross-frequency features achieved an AUC of 0.653 (95 % CI: 0.45-0.70), with alpha/beta nodal metrics (e.g., O2α2: absolute standardized β >0.5).

By systematically mapping brain region-frequency specific network anomalies, we validated a neurophysiological model of depressive symptoms characterized by multi-frequency, multiregional dysregulation. These oscillatory signatures provide electrophysiological targets for personalized neuromodulation, with model interpretability bridging symptom-network dynamics and neurocircuit mechanisms.