Decoding cortical folding patterns in marmosets using machine learning and large language model.

Area of Science:

• Neuroscience
• Genomics
• Computational Biology

Background:

• Macroscale neuroimaging shows differences in primate cerebral cortex connectivity between gyri and sulci.
• Understanding these differences at a molecular level remains a significant challenge.

Purpose of the Study:

• To systematically analyze cortical folding patterns at the molecular level using in situ hybridization data.
• To identify genes with transcriptomic differences between sulci and gyri using machine learning and large language models (LLMs).

Main Methods:

• Utilized a comprehensive dataset of whole-brain in situ hybridization (ISH) data from marmosets.
• Applied advanced machine learning algorithms and a large language model (LLM) to analyze transcriptomic data.
• Performed gene enrichment, neural migration, and axon guidance pathway analyses.

Main Results:

• Identified specific genes exhibiting significant transcriptomic differences between concave (sulci) and convex (gyri) cortical patterns.
• Elucidated molecular mechanisms potentially underlying structural and functional differences in cortical folding.
• Demonstrated the utility of LLMs in analyzing complex neurobiological datasets.

Conclusions:

• Provides novel insights into the molecular basis of cortical folding in the primate brain.
• Highlights the potential of integrating LLMs with transcriptomic data for understanding brain structure and function.
• Establishes a foundation for future research into the genetic regulation of brain morphology.