A data-driven framework for mapping domains of human neurobiology

Area of Science:

• Neuroscience
• Computational Biology
• Medical Imaging

Background:

• Functional neuroimaging, particularly functional magnetic resonance imaging (fMRI), has been crucial in neuroscience for 25 years.
• Current interpretation of fMRI data relies on expert-defined frameworks, which can introduce biases and limit research replicability.
• There is a need for more objective, data-driven approaches to understand brain structure-function relationships.

Purpose of the Study:

• To develop a data-driven computational framework for neurobiological domains using a large corpus of neuroimaging literature.
• To improve the replicability of structure-function links derived from fMRI data.
• To create a comprehensive, quantifiable characterization of human brain circuits.

Main Methods:

• A computational approach was used to synthesize text and data from nearly 20,000 human neuroimaging articles.
• A data-driven framework was derived across multiple levels of domain specificity.
• The framework's ability to replicate structure-function links was tested against held-out articles.

Main Results:

• The data-driven framework demonstrated better replication of structure-function links compared to dominant expert-defined frameworks.
• The framework successfully partitioned the literature into modular subfields.
• Domains within the framework served as generalizable prototypes for structure-function patterns in individual studies.

Conclusions:

• The developed computational ontology provides a comprehensive and quantifiable characterization of human brain circuits using fMRI data.
• This data-driven approach enhances the objectivity and replicability of neuroimaging research.
• The methodology holds potential for synthesizing other scientific literatures beyond neuroimaging.