A multimodal approach for visualizing and identifying electrophysiological cell types in vivo

Area of Science:

• Neuroscience
• Computational Biology
• Data Science

Background:

• Electrophysiological recordings capture neural activity but struggle with precise cell type identification.
• Differentiating neuron types is crucial for understanding neural circuit dynamics and function.

Purpose of the Study:

• To introduce PhysMAP, a novel framework for simultaneous multi-cell type identification using electrophysiological data.
• To enhance the accuracy and interpretability of cell type classification in neural recordings.

Main Methods:

• PhysMAP adapts multiomics analysis techniques to integrate multiple electrophysiological modalities.
• It generates interpretable multimodal representations by weighting different data types.
• The framework was validated across seven diverse electrophysiological datasets.

Main Results:

• Multimodal representations generated by PhysMAP show superior alignment with transcriptomically-defined cell types compared to single modalities.
• PhysMAP accurately identifies putative cell types even without ground truth labels.
• The framework successfully transfers labels between annotated and unannotated datasets and detects batch effects.

Conclusions:

• PhysMAP provides a robust tool for simultaneous multi-cell type analysis in neuroscience.
• This framework offers deeper insights into neural circuit dynamics by improving cell type identification.
• PhysMAP enhances the reliability of neural data analysis by mitigating batch effects.