Decoding Depression: Integrating Brain Networks and Blood Biomarkers in Teens

Ding Zhou¹QI ZHANG¹Wanying Qiao¹Xinzhu Zheng²Min Zhang²Jinhui Zhang²Ying Li²Lili Fan^1*Jin-yu Wu²Liping Liu¹

• ¹Harbin Medical University, harbin, China
• ²The first Psychiatric Hospital of Harbin, harbin, China

Functional near-infrared spectroscopy (fNIRS) is a neuroimaging technique without invasion that has been applied to assess hemodynamic changes in the cerebral cortex. However, systematic studies on the interaction between brain network function and blood biomarkers are lacking, especially in adolescents. This study recruited 207 adolescents (aged 12-18) categorized into four groups based on HAMD-17 scores. Following Permutation test and Holm correction, channel-specific beta (β) values indicated cortical activation. Task-based fNIRS revealed significantly reduced activation in the depression group versus controls within prefrontal (Ch7, Ch15) and temporal (Ch33) regions. Functional connectivity matrices demonstrated progressively decreasing connectivity in depression groups with increasing symptom severity. Graph theoretical analysis across sparsity thresholds (0.05-0.6) showed the depression group had significantly lower clustering coefficient, small-worldness, local efficiency, cycle probability, and global efficiency. Spearman correlation and Partial correlation indicated that high density lipoprotein cholesterol were significant positive correlations with channel Eloc data (Ch33), after FDR correction. . In path analysis, the model fit was excellent (p = 0.817, RMSEA = 0.000, SRMR = 0.017, CFI = 1.000, R² = 0.16). HDL-C was the strongest predictor of depression scores (β = 0.293, p < 0.001), and mode of residence exerted a significant positive influence (β = 0.178, p = 0.006). This study combined fNIRS with blood multi-omics data, revealing reduced activation in channels 7, 15, and 33 among depression patients, accompanied by decreased C and Eloc levels. Concurrently, HDL-C was found to correlate significantly with depression severity, and an effective predictive model for depression was constructed.