Study on Quantitative Expression and Treatment Response of Latent Disease Factors Underlying Psychopathological Dimensions Revealed by Unsupervised Machine Learning

Research Background

Heterogeneity and comorbidity are prevalent in psychiatric diagnoses, posing challenges for precise diagnosis and personalized treatment. For instance, Autism Spectrum Disorder (ASD), Attention Deficit/Hyperactivity Disorder (ADHD), and Obsessive-Compulsive Disorder (OCD) often overlap in symptom diagnosis. Their co-occurring symptoms may be mediated by shared and/or different neural mechanisms, but are difficult to categorize at the individual level. The application of advanced Bayesian models and unsupervised machine learning techniques provides a quantitative and individualized approach to analyze the relationship between psychopathological dimensions and disease factors.

Paper Information

• Authors: Zhao Shaoling, Lü Qian, Zhang Ge, Zhang Jiangtao, Wang Heqiu, Zhang Jianmin, Wang Meiyun, Wang Zheng
• Affiliations: Institute of Neuroscience, Chinese Academy of Sciences; Department of Psychology, Peking University; Department of Medical Imaging, Henan Provincial People’s Hospital; Tongde Hospital of Zhejiang Province (Zhejiang Mental Health Center)
• Publication Date and Venue: Accepted on January 2, 2024, published in “Neuroscience Bulletin”
• Research Type: Original research

Detailed Research Process

a) Research Flow

The research design was divided into several stages. First, researchers extracted latent disease factors from a mixed cohort of ASD and ADHD using unsupervised learning and hierarchical Bayesian framework. Further, canonical correlation analysis was used to explore the relationship between individual performance dimensions and latent factors. The model was extended to external validation of unseen individuals based on the same factors, including subclinical populations and local OCD databases.

b) Research Results

The study identified four main latent disease factors, which showed significant correlations with different symptom domains. These factors were variably expressed in individuals and could significantly predict individual symptom scores and treatment responses. The expression level of predictive factors was associated with the degree of symptom improvement in OCD patients.

c) Conclusions and Significance

The study demonstrates that data-derived latent disease factors can quantify individual factor expression, informing dimensional symptoms and treatment outcomes across cohorts, which helps promote quantitative diagnosis and personalized intervention for mental disorders.

d) Research Highlights

• Novel methodology: Advanced Bayesian models and unsupervised machine learning techniques were used to analyze brain functional connectivity data.
• Multidimensional factors: The study not only revealed the variable expression of latent disease factors among individuals but also validated the universality and predictive value of these factors through external datasets.
• Clinical value: The results can precisely indicate the treatment response of OCD patients, which is of great significance for clinical evaluation of treatment efficacy.

Conclusion

By applying advanced machine learning techniques, this study provides a new quantitative interpretation for dimensional diagnosis and clinical intervention of mental disorders, offering important guidance and inspiration for future research on precision psychiatry.