Non-linear Dimensionality Reduction on Extracellular Waveforms Reveals Cell Type Diversity in Premotor Cortex

Abstract

Cortical circuits involved in decision-making are thought to contain a large number of cell types—each with different physiological, functional, and laminar distribution properties—that coordinate to produce behavior. Current in vivo methods rely on clustering of specified features, such as trough to peak duration of extracellular spikes, to identify putative cell types these but can only capture a small amount of variation. Here, we develop a new method (WaveMAP) that combines non-linear dimensionality reduction with graph clustering to identify putative cell types. We apply WaveMAP to extracellular waveforms recorded from dorsal premotor cortex of macaque monkeys performing a decision-making task. Using WaveMAP, we robustly establish eight waveform clusters and show that these clusters recapitulate previously identified narrow- and broad-spiking types while also revealing undocumented diversity. These clusters exhibit distinct laminar distributions, characteristic firing rate patterns, and decision-related dynamics. By revealing additional cell type diversity, WaveMAP facilitates a more nuanced understanding of how the dynamics of cell types unfolds across cortical layers during decision-making.

Large-scale neural recordings with single-cell resolution in human cortex using high-density Neuropixels probes

“…We also utilized an automated, non-linear method for classifying extracellular waveforms, with the goal of identifying additional classes based on smaller and more nuanced features of the waveforms 51 . Using the WaveMAP algorithm 51 , we found that certain waveform types were identifiable across patients. Negative RS and FS-like units were present in all three patient’s data, while the positive large waveforms (PS) appeared in two of the three cases ( Fig.…”

“…Similar waveform clusters were found across participants using each of the three separate classification approaches, indicating there were consistencies between individuals. This waveform diversity presumably relates to both differences in cell types and differences in where electrical activity is recorded (e.g., near soma versus axons) 51 .…”

“…We found that, within minutes after inserting the Neuropixels probe into cortical tissue, we can observe 10s to more than 200 clear and stable units. We could separate units based on waveform shapes (using duration and polarity), and we could also cluster the waveforms using unbiased clustering via PCA with k-means clustering and, separately, using a novel nonlinear clustering approach, WaveMAP 51 . Similar waveform clusters were found across participants using each of the three separate classification approaches, indicating there were consistencies between individuals.…”

P-sort: an open-source software for cerebellar neurophysiology

“…It is made The copyright holder for this preprint this version posted March 17, 2021. ; https://doi.org/10.1101/2021.03.16.435644 doi: bioRxiv preprint inter-connected in the UMAP space. This approach was recently demonstrated by Lee et al (2021) in sorting of cortical spikes.…”

“…This has the unique advantage of allowing for cross validation, which can be employed by semi-supervised learning methods in which the expert labels a subset of the data and leaves it to UMAP to make predictions on the unlabeled data set. Indeed, UMAP was recently adopted by Markanday et al (2020) to cluster complex spikes from the cerebellum, and by Lee et al (2021) to cluster spikes from the cerebral cortex.…”