Identifying shared and personalized brain functional connectivity subspace across neuropsychiatric disorders.
Where this comes from
- Record sourced from PubMed, PMID 42710528.
- Also identified by DOI 10.1088/1741-2552/aea44a.
- No licence information is recorded for this record.
- Because redistribution is not established, this page shows the abstract only. Follow the links below for the full text.
Abstract
Neuropsychiatric disorders exhibit complex functional connectivity (FC) alterations, yet disentangling transdiagnostic mechanisms from disorder-specific network perturbations remains challenging due to clinical heterogeneity and multisite data confounds. 

Approach: We aggregated large-scale resting-state fMRI data from 1,923 patients with Major Depressive Disorder (MDD), Autism Spectrum Disorder (ASD), or Attention-Deficit/Hyperactivity Disorder (ADHD), and site-matched controls from the REST-meta-MDD, ABIDE, and ADHD-200 consortia. We applied the established Common Orthogonal Basis Extraction (COBE) algorithm to decompose individual FC matrices into shared and personalized subspaces following site-level normalization.

Main results: We validated that shared subspaces captured robust pathological signatures capable of distinguishing patients from independent healthy controls from the Human Connectome Project (HCP). A hierarchical analysis of these shared subspaces revealed a transdiagnostic core defined by widespread default mode network (DMN) decoupling alongside divergent subcortical connectivity patterns: hyper-connectivity in MDD but hypo-connectivity in ASD and ADHD. Despite these conserved commonalities, we demonstrated a functional dissociation in clinical utility: personalized subspaces exhibited significantly higher utility for precision characterization, achieving superior performance in predicting individual symptom severity and in discriminating between diagnostic groups (macro-F1: 77.6%)- tasks where shared features lacked sufficient specificity. 

Significance: These findings provide evidence consistent with a hierarchical neurobiological architecture wherein shared subspaces map conserved vulnerabilities, while personalized subspaces capture the heterogeneity underlying diagnostic distinctions and individual symptom expression. This work supports the value of modeling personalized neural signatures to advance precision psychiatry.
.