Asymmetric Branching Scale Factors as Features in Neuronal and Glial Cell-Type Classification Using Machine Learning Methods.

Neurons are connected by complex branching processes - axons and dendrites - that process information for organisms to respond to their environment. Classifying neurons according to differences in structure or function is a fundamental piece of neuroscience. In previous work, we constructed a biophysical theory that establishes a correspondence between neuron structure and function as mediated by principles such as time or power minimization, using undetermined Lagrange multipliers to predict scaling ratios for axon and dendrite sizes across branching levels. Here, we relax the assumption of symmetrical branching in the model to determine asymmetric branching powers that differ across different cell types due to functional tradeoffs. Furthermore, we use scale factors related to asymmetric branching as features in machine learning classification to distinguish between different cell types. We find significant distinctions in the asymmetric scaling ratios between Purkinje cells and motoneurons and between axons and microglia, a specific class of electrically active non-neuronal brain cells. The performance of these classification methods gives us important insights into the correspondence between structure and function across different cell types.