An integrated multimodal model of alcohol use disorder generated by data-driven causal discovery analysis

Area of Science:

• Neuroscience
• Psychiatry
• Data Science

Background:

• Alcohol use disorder (AUD) poses significant public health challenges.
• Limited understanding of the complex neurobehavioral mechanisms underlying AUD.
• Need for studies investigating causal pathways to AUD severity.

Purpose of the Study:

• To analyze causal pathways to AUD severity using Causal Discovery Analysis (CDA).
• To model relationships between phenotypic factors, brain connectivity, and AUD.
• To identify key neurobehavioral drivers of AUD.

Main Methods:

• Utilized data from the Human Connectome Project (HCP; n=926).
• Applied exploratory factor analysis to 100 phenotypic measures.
• Assessed functional connectivity in 12 resting-state brain networks.
• Employed data-driven CDA to generate a causal model.

Main Results:

• A causal model was generated linking brain connectivity, cognition, social, and affective factors to AUD severity.
• Causal influence hierarchy identified: brain connectivity → cognition → social → affective/psychiatric function → AUD severity.
• Confirmed hypothesized cognitive and affective influences, highlighting the importance of social factors.

Conclusions:

• Neurobehavioral factors, including brain connectivity, cognition, social, and affective domains, causally influence AUD severity.
• Current addiction models should be expanded to incorporate the significant role of social factors.
• Data-driven causal modeling provides novel insights into AUD etiology.