Predictive models for social functioning in healthy young adults: A machine learning study integrating...

Area of Science:

• Neuroscience
• Psychology
• Public Health

Background:

• Poor social functioning is a growing public health concern with significant physical and mental health implications.
• Prognostic tools are essential for identifying at-risk individuals and guiding timely interventions for social functioning deficits.

Purpose of the Study:

• To evaluate machine learning models for predicting social functioning using bio-behavioral data.
• To determine the predictive accuracy of models incorporating brain morphology, cognition, and behavior.

Main Methods:

• Utilized Support Vector Regression (SVR) models on data from the Human Connectome Project (N=1,101, ages 22-35).
• Compared four models: brain-only, brain-cognition, cognition-behavioral, and combined brain-cognition-behavioral.
• Assessed predictive accuracy and identified key predictor variables.

Main Results:

• The combined (brain-cognition-behavior) and cognition-behavioral models significantly predicted social functioning.
• Brain-only and brain-cognition models showed no significant predictive power.
• Key predictors included negative affect, psychological wellbeing, extraversion, withdrawal, and specific cortical thickness regions.

Conclusions:

• Machine learning accurately predicts social functioning in healthy young adults.
• Behavioral markers appear more influential than brain measures alone for predicting social functioning.
• These findings highlight behavioral factors as crucial targets for preventative interventions.