A Bayesian Regularized and Annotation-Informed Integrative Analysis of Cognition (BRAINIAC)

Area of Science:

• Neuroscience
• Computational Biology
• Psychiatry

Background:

• Understanding brain-behavior relationships is crucial for cognitive phenotyping.
• Existing models often assume sparse associations, which may not reflect biological complexity.
• Integrating biological annotations can enhance predictive models of cognition.

Purpose of the Study:

• Introduce the Bayesian Regularized and Annotation-Informed Integrative Analysis of Cognition (BRAINIAC) model.
• Estimate total variance explained by features for cognitive phenotypes.
• Assess the impact of annotations on feature enrichment without assuming sparsity.

Main Methods:

• Developed the BRAINIAC model for integrative analysis of neuroimaging and cognitive data.
• Validated the model using Monte Carlo simulations.
• Applied BRAINIAC to resting-state functional magnetic resonance imaging (rsMRI) and neuropsychiatric data from the Adolescent Brain Cognitive Development (ABCD) Study.

Main Results:

• BRAINIAC accurately estimates variance explained by features for cognitive phenotypes.
• The model demonstrates improved out-of-study predictive power when incorporating relevant annotations.
• Validation studies confirmed the robustness of the BRAINIAC approach.

Conclusions:

• The BRAINIAC model offers a novel framework for analyzing brain-behavior associations.
• Annotation-informed analysis significantly enhances predictive accuracy for cognitive phenotypes.
• This approach advances our understanding of the neural underpinnings of cognition.