Refining RDoC Using Individual-Level Task fMRI Factor Models Reveals Reproducible and Clinically Relevant Brain-Wide...

Area of Science:

• Neuroscience
• Psychiatry
• Computational Biology

Background:

• The Research Domain Criteria (RDoC) framework aims to structure psychiatric research using biologically grounded, dimensional constructs.
• The hierarchical structure of the RDoC framework lacks validation against individual-level brain and behavioral data.

Purpose of the Study:

• To test whether individual-level, data-driven models more accurately capture brain activity and behavior organization than RDoC-based models.
• To validate and refine dimensional psychiatric frameworks using empirical data.

Main Methods:

• Applied a multi-stage validation framework to Human Connectome Project (HCP) task-fMRI data.
• Utilized confirmatory factor analysis, community detection, and topological analysis on neural and behavioral data from HCP and LA5c datasets.
• Included healthy controls and individuals with ADHD, bipolar disorder, and schizophrenia.

Main Results:

• Data-driven bifactor models consistently outperformed RDoC-based models in capturing brain activity and behavior organization across independent cohorts.
• Data-driven models revealed reproducible spatial motifs and a general factor aligning with cortical gradients, predicting task performance.
• Data-driven models generalized across diagnostic groups and related to symptom severity, outperforming RDoC representations.

Conclusions:

• Individual-level, empirically derived models provide a more accurate, generalizable, and behaviorally relevant account of brain organization than the RDoC framework.
• This work advances precision neuroscience by integrating neural, behavioral, and clinical validation.
• Findings support the empirical refinement of dimensional psychiatric frameworks for improved mental health research and treatment.