Manifold Learning Uncovers Nonlinear Interactions Between the Adolescent Brain and Environment That Predict Emotional...

Area of Science:

• Neuroscience
• Computational Biology
• Developmental Psychology

Background:

• Adolescent mental health research requires methods to link brain development with environmental experiences.
• Current understanding of neurobiological embedding in the environment is limited.
• Developing new approaches is crucial for advancing adolescent mental health research.

Purpose of the Study:

• To investigate associations between brain function, environments, and emotional/behavioral problems in adolescents.
• To develop and apply a novel computational method (E-PHATE) for modeling brain-environment interactions.
• To predict individual differences in cognition and emotional/behavioral problems using brain-environment data.

Main Methods:

• Utilized data from the Adolescent Brain Cognitive Development (ABCD) Study (n=2401 female).
• Applied manifold learning, specifically developing exogenous PHATE (E-PHATE), to analyze high-dimensional functional magnetic resonance imaging data.
• E-PHATE embeddings of brain activation and environment were used to predict cross-sectional and longitudinal outcomes.

Main Results:

• E-PHATE embeddings demonstrated moderate-to-large associations with emotional and behavioral problems at baseline, outperforming traditional methods.
• Baseline E-PHATE embeddings predicted emotional and behavioral problems two years later.
• Longitudinal predictions showed consistent moderate effects in frontoparietal and attention networks.

Conclusions:

• The embedding of the adolescent brain within its environment provides deeper insights into emotional and behavioral problems.
• E-PHATE harmonizes advanced computational techniques with developmental theories.
• This approach enhances the detection and prediction of adolescent emotional and behavioral problems.