The use of cluster analysis for cell typing

In invertebrates, large neurons, identifiable in each animal, have proven to be useful models for investigating basic neurobiological phenomena. In vertebrates, the number of neurons and the complexity of nervous systems increase and 'identifiability' is lost. To compensate for this, other approaches must be adopted to study the vertebrate brain. One successful approach has been to identify cell types, recognizable in each individual. The identification of cell types in central nuclei has helped us understand the organization of these nuclei and has provided an important foundation for examining possible relationships between the structure and function of neurons. Unfortunately, not all nuclei are composed of neurons of readily identifiable types. Nuclei lacking distinct cell types are, in general, less well understood than nuclei with morphologically distinct cell types. This article describes a statistical approach known as cluster analysis that we [11] used to define cell types in the nucleus of the solitary tract - a nucleus that had been suggested to contain identifiable cell types [3,4,14,18] but within which cell typing had proven difficult. Similar techniques have been used to classify Golgi-impregnated cells in the ventrobasal complex of the dog [5], the subthalamic nucleus of the bushbaby [9] and the human retina [6]. Cluster analysis has also been used in the gustatory system [10,16], in general, and nucleus of the solitary tract, in specific, to classify electrophysiologically characterized cells [2,8,15]. The method also includes a technique for verifying the utility of the resulting classification.