Classification accuracy of structural and functional connectomes across different depressive phenotypes

Phenotyping of major depressive disorder (MDD) can vary from study to study, which, together with heterogeneity of the disorder, may contribute to the inconsistent associations with neuroimaging features and underlie previous problems with machine-learning methods for MDD diagnostic applications. In this study, we examined the classification accuracy of structural and functional connectomes across different depressive phenotypes, including separating MDD subgroups into those with and without self-reported exposure to childhood trauma (CT) (one of the largest risk factors for MDD associated with brain development). We applied logistic ridge regression to classify control and MDD participants defined by six different MDD definitions in a large community-based sample ( ). We used brain connectomic data based on six structural and two functional network weightings and conducted a comprehensive analysis to (i) explore how well different connectome modalities predict different MDD phenotypes commonly used in research, (ii) whether stratification of MDD based on self-reported exposure to childhood trauma (measured with the childhood trauma questionnaire (CTQ)) may improve the accuracies, and (iii) identify important predictive features across different MDD phenotypes. We found that functional connectomes outperformed structural connectomes as features for MDD classification across phenotypes. The highest accuracy of 64.8% (chance level 50.0%) was achieved in the Currently Depressed (defined by the presence of more than five symptoms of depression in the past 2 weeks) sample with additional CTQ criterion using partial correlation functional connectomes. The predictive feature overlap, measured using Jaccard index, indicated that there were neurobiological differences between MDD patients with and without childhood adversity. Further analysis of predictive features for different MDD phenotypes with hypergeometric tests revealed sensorimotor and visual subnetworks as important predictors of MDD. Our results suggest that differences in sensorimotor and visual subnetworks may serve as potential biomarkers of MDD.