Detecting cell assemblies by NMF-based clustering from calcium imaging data

Area of Science:

• Neuroscience
• Computational Biology
• Data Science

Background:

• Neurons form cell assemblies crucial for brain information processing.
• Advancements in calcium imaging enable studying these neural networks.
• Extracting meaningful patterns from complex neural data remains a challenge.

Purpose of the Study:

• To develop and validate a novel clustering approach for extracting cell assemblies from calcium imaging data.
• To address the model selection problem in non-negative matrix factorization (NMF).
• To investigate the relationship between cell assembly size and cortical slow-wave activity.

Main Methods:

• Utilized non-negative matrix factorization (NMF) to derive a neuron similarity matrix.
• Applied spectral clustering on the NMF-derived similarity matrix.
• Introduced a model averaging technique with a novel estimator to resolve NMF model selection issues.
• Validated the approach using simulated data and real calcium imaging data from mice.

Main Results:

• The proposed NMF-based clustering method demonstrated superior performance compared to correlation-based methods across various sampling rates in simulated data.
• Analysis of mouse calcium imaging data revealed a correlation between cell assembly size and the degree/spatial extent of cortical slow-wave generation.
• The model averaging approach effectively addressed the NMF model selection challenge.

Conclusions:

• The NMF-based spectral clustering offers a robust method for cell assembly extraction from calcium imaging.
• Cell assembly dynamics are influenced by cortical slow-wave activity during different behavioral states (sleep/wake).
• The developed method provides a valuable tool for systems neuroscience research.