Reliable Inference of the Encoding of Task States by Individual Neurons Using Calcium Imaging.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Calcium imaging is widely used to measure neuronal activity for studying the neural basis of behavior.
• Diverse methodological choices in data analysis can affect the inferred selectivity of neurons.
• Lack of standardized parameters hinders reliable assessment of neuronal encoding preferences.

Purpose of the Study:

• To systematically examine the impact of parameter choices on neuronal selectivity inference from calcium imaging data.
• To identify an optimal combination of parameters for accurate and robust assessment of neuronal encoding.
• To establish a generalizable data-driven procedure for optimizing analysis pipelines.

Main Methods:

• Calcium imaging in the medial prefrontal cortex of freely behaving mice during an exploration-avoidance task.
• Systematic variation of parameters across data acquisition, processing, and statistical testing.
• Development and application of novel accuracy and robustness metrics for parameter evaluation.
• Validation using resampling methods and comparison across different analytical approaches (selectivity indices, regression coefficients).

Main Results:

• Key parameters significantly impact inferred neuronal selectivity in distinct ways.
• An optimal parameter combination was identified, yielding accurate and robust inferences.
• The optimal settings were validated and shown to be generalizable across analytical methods.

Conclusions:

• Identified an optimal parameter setting for reliable assessment of encoding preferences in cortical excitatory neurons using GCaMP6f calcium imaging.
• The optimal combination includes specific filtering, data binning, dataset sizing, and shuffling methods.
• Established a generalizable data-driven procedure for optimizing calcium imaging analysis pipelines for various cell types, brain regions, and tasks.