Three Latent Factors in Major Depressive Disorder Base on Functional Connectivity Show Different Treatment...

Area of Science:

• Neuroscience
• Psychiatry
• Computational Biology

Background:

• Major depressive disorder (MDD) is highly heterogeneous, complicating treatment selection.
• Existing subtype classifications often neglect individual variability within groups.
• Resting-state functional connectivity (FC) offers a potential avenue for understanding MDD heterogeneity.

Purpose of the Study:

• To decompose resting-state FC into latent factors using latent dirichlet allocation (LDA) to capture inter-individual variability in MDD.
• To identify distinct patterns of hyper- and hypo-connectivity associated with these factors.
• To investigate the relationship between these connectivity patterns and treatment preferences in MDD patients.

Main Methods:

• Latent Dirichlet Allocation (LDA) applied to resting-state functional connectivity (FC) data from 226 MDD patients and 100 healthy controls.
• Identification of latent factors representing unique FC compositions.
• Examination of associations between identified factors, connectivity patterns, demographic data, clinical symptoms, and treatment outcomes.

Main Results:

• Three distinct FC factors were identified.
• Factor 1 (Default Mode Network [DMN] hyperconnectivity) associated with antidepressant monotherapy response, younger age, higher education, and cognitive symptom improvement.
• Factor 3 (DMN hypo-connectivity) associated with combined antidepressant and stimulation therapy response; Factor 2 (global hypo-connectivity) linked to higher depression and anxiety severity.

Conclusions:

• Distinct functional connectivity patterns in MDD correlate with specific clinical characteristics and predict differential treatment responses.
• Patients with DMN hyperconnectivity may benefit from monotherapy, while those with DMN hypo-connectivity may respond better to combined treatments.
• This factor-based approach offers a novel strategy for developing personalized biomarkers for MDD treatment selection.