Accurate Estimation of Neural Population Dynamics without Spike Sorting.

Area of Science:

• Systems Neuroscience
• Computational Neuroscience
• Neurophysiology

Background:

• Relating neural activity to behavior is a core objective in systems neuroscience.
• Analyzing neural population dynamics often involves reducing data dimensionality.
• Spike sorting is a critical but increasingly challenging step in neural data analysis, especially with large datasets.

Purpose of the Study:

• To investigate whether spike sorting is a necessary step for estimating neural population dynamics.
• To explore the applicability of random projection theory to neural data analysis.
• To determine if simplified data processing methods yield comparable scientific conclusions.

Main Methods:

• Applied random projection theory to analyze neural population activity.
• Recorded neural data using Neuropixels probes in the motor cortex of nonhuman primates.
• Reanalyzed data from three independent prior studies.
• Compared results obtained using multiunit threshold crossings versus sorted single neurons.

Main Results:

• Neural population dynamics and scientific conclusions were found to be highly similar when using multiunit threshold crossings instead of sorted single neurons.
• The geometry of low-dimensional manifolds could be accurately estimated from linear projections of neural data.
• The findings support the hypothesis that complex spike sorting may be bypassed for certain analyses.

Conclusions:

• Spike sorting may not be strictly necessary for estimating neural population dynamics and drawing scientific conclusions.
• This simplification can unlock vast amounts of existing neural data for new analyses.
• The findings have implications for designing future electrode arrays and optimizing laboratory and clinical data acquisition.