White-matter functional topology: a neuromarker for classification and prediction in unmedicated depression

Abstract

Aberrant topological organization of brain connectomes underlies pathological mechanisms in major depressive disorder (MDD). However, accumulating evidence has only focused on functional organization in brain gray-matter, ignoring functional information in white-matter (WM) that has been confirmed to have reliable and stable topological organizations. The present study aimed to characterize the functional pattern disruptions of MDD from a new perspective—WM functional connectome topological organization. A case-control, cross-sectional resting-state functional magnetic resonance imaging study was conducted on both discovery [91 unmedicated MDD patients, and 225 healthy controls (HCs)], and replication samples (34 unmedicated MDD patients, and 25 HCs). The WM functional networks were constructed in 128 anatomical regions, and their global topological properties (e.g., small-worldness) were analyzed using graph theory-based approaches. At the system-level, ubiquitous small-worldness architecture and local information-processing capacity were detectable in unmedicated MDD patients but were less salient than in HCs, implying a shift toward randomization in MDD WM functional connectomes. Consistent results were replicated in an independent sample. For clinical applications, small-world topology of WM functional connectome showed a predictive effect on disease severity (Hamilton Depression Rating Scale) in discovery sample (r = 0.34, p = 0.001). Furthermore, the topologically-based classification model could be generalized to discriminate MDD patients from HCs in replication sample (accuracy, 76%; sensitivity, 74%; specificity, 80%). Our results highlight a reproducible topologically shifted WM functional connectome structure and provide possible clinical applications involving an optimal small-world topology as a potential neuromarker for the classification and prediction of MDD patients.