Preliminary Phenotypic Cluster Analysis of Cardiorespiratory Modulated Neuronal Discharge Patterns with Dynamic Visualizations

Abstract

Multi‐dimensional cluster analysis has been applied successfully to many taxonomy problems. Building on previous work (see Carroll et al., 2013 & Zuperku et al., 2015), we have developed a quantitative, reproducible clustering classification of respiratory neuron phenotypes based upon activity. We displayed the results with a multi‐factorial visualization tool.We curated 100‐bin, normalized, cycle triggered histograms of spike trains with significant respiratory modulation (n=573 from a library of 1,373), recorded in lateral respiratory column (ventral respiratory group, pre‐Bötzinger, Bötzinger), pons, raphé, and dorsal lateral medulla. Initial analysis by Ward clustering with Cartesian distance in 100‐space differentiated three groups: predominantly inspiratory, predominantly expiratory and low respiratory modulation.Subsequent application of the same clustering paradigm to each of these 3 groups yielded a preliminary set of 24 archetypes, the averaged histograms of the elements of each cluster. We used those archetypes as clustering centroids on naïve recordings of 229 spike trains; all were clustered, with 20 of the archetypes. A 3‐D brainstem atlas displayed neuronal firing rates during periods of control and/or stimulated activity such as chemoreception, cough, swallow and vagotomy; cell activity was colored by type (cluster), illuminated in proportion to firing rate, and could be displayed in stereo or as a side‐by‐side or differential comparisons of stimulated vs. control activity.The data do not support a continuum of phenotypes based on discharge identity but instead are consistent with definable phenotypes.Support or Funding InformationOT2OD023854, HL109025This abstract is from the Experimental Biology 2018 Meeting. There is no full text article associated with this abstract published in The FASEB Journal.