Facilitating analysis of open neurophysiology data on the DANDI Archive using large language model tools

Area of Science:

• Neuroscience
• Data Science
• Computational Biology

Background:

• The DANDI Archive hosts over 400 open neurophysiology datasets in the Neurodata Without Borders (NWB) format.
• Researchers face challenges in accessing and identifying relevant neurophysiology data for reuse due to unfamiliarity with methods and content.
• Facilitating data reuse is crucial for advancing scientific discovery and adhering to FAIR data principles.

Purpose of the Study:

• To develop and evaluate AI-powered tools to lower barriers to accessing and reusing neurophysiology data in the DANDI Archive.
• To enhance user engagement with open neurophysiology datasets through automated assistance and guided analysis.

Main Methods:

• An AI-powered, agentic chat assistant was developed to guide users through data exploration, access, visualization, and preliminary analysis.
• A notebook generation pipeline was created to automatically analyze dataset structures, execute inspection scripts, and generate visualizations.
• The system was applied to 12 recent DANDI datasets, and the generated notebooks were reviewed by neurophysiology data specialists.

Main Results:

• The AI chat assistant and notebook generator were successfully applied to 12 neurophysiology datasets.
• Generated Python notebooks were found to be generally accurate and well-structured by data specialists.
• The majority of reviewed notebooks were rated as "very helpful" by neurophysiology data specialists.

Conclusions:

• AI-powered tools can significantly lower barriers to reusing open neurophysiology data.
• Automated analysis and guided exploration enhance data accessibility and engagement, supporting FAIR data principles.
• This approach demonstrates the potential of AI to streamline data analysis workflows in neuroscience research.