Neuron subtypes play the long game

Abstract

Approaches to assigning neurons to specific subtypes in the mouse brain are commonly based on electrophysiological, morphological and transcriptomic properties. Connectomic analyses for subtype classification have typically been restricted to electron microscopy datasets, but the authors of a study in Nature Methods instead leverage high-throughput single-neuron reconstruction datasets to classify cells on the basis of potential connectivity. The authors compiled a database of over 20,000 single-neuron morphological reconstructions and used machine learning to assign topologically connected axonal and dendritic arbors. Potential connectivity was therefore obtained from the overlap of dendritic and axonal arbor spatial domains. Importantly, classification of cells based on connectivity, or c-types, outperformed classification based on morphology, or conventional m-types, and substantially contributed to classification performance jointly with m-types. Further subtyping was possible using spatially correlated connectivity and morphological features. Using c-type classification, the authors were able to showcase subtyping of thalamic and thalamocortical neurons. Potential connectivity subtyping is thus a promising method by which to classify individual neurons in the brain into types.

Original reference: Nat. Methods https://doi.org/10.1038/s41592-025-02621-6 (2025)