Related Experiment Video
Updated: Aug 14, 2026
09:44
Recording and Analyzing Multimodal Large-Scale Neuronal Ensemble Dynamics on CMOS-Integrated High-Density Microelectrode Array
Published on: March 8, 2024
Constructing mesoscale functionomics by neural dynamics subspace clustering
Yeyi Cai1,2, Xinhong Xu1,2, Guihua Xiao1,2,3
1Department of Automation, Tsinghua University, Beijing 100084, China.
National Science Review
|August 13, 2026
Summary
We developed Functional subspace clustering based on sparse Representation of Intrinsic Dynamics (FRID) to identify neurons with shared properties. FRID improves brain function mapping and understanding neural circuit dynamics.
Area of Science:
- Neuroscience
- Computational Biology
- Data Science
Background:
- Mapping brain function at a fine-grained level is essential for understanding intelligence.
- Current methods struggle with complex neural interactions and asynchronous firing patterns.
Purpose of the Study:
- To introduce Functional subspace clustering based on sparse Representation of Intrinsic Dynamics (FRID), an unsupervised method.
- To identify neurons with shared microcircuit connectivity and information encoding properties ('Functionomics') from neural recordings.
Main Methods:
- FRID utilizes sparse representation of intrinsic neural dynamics.
- The approach is unsupervised, requiring no prior labels.
- Applied to both simulated complex networks and empirical calcium recordings.
Main Results:
- FRID outperforms traditional correlated-firing-based clustering methods.
- Achieved higher resolution and coding specificity in functionomic clusters compared to anatomical areas.
- Demonstrated functional remapping after ischemic stroke and learning-induced reorganization.
Conclusions:
- FRID offers a robust method for identifying neuronal function ('Functionomics').
- The approach enhances the understanding of mesoscale brain functions and neural circuit dynamics.
- FRID serves as a general data-driven paradigm for neuroscience research.
