Brains are not engineered solutions to a well-defined problem but arose through selective pressure acting on random variation. It is therefore unclear how well a model chosen by an experimenter can relate neural activity to experimental conditions. Here we developed 'Model identification of neural encoding (MINE)'. MINE is an accessible framework using convolutional neural networks (CNN) to discover and characterize a model that relates aspects of tasks to neural activity. Although flexible, CNNs are difficult to interpret. We use Taylor decomposition approaches to understand the discovered model and how it maps task features to activity. We apply MINE to a published cortical dataset as well as experiments designed to probe thermoregulatory circuits in zebrafish. MINE allowed us to characterize neurons according to their receptive field and computational complexity, features which anatomically segregate in the brain. We also identified a new class of neurons that integrate thermosensory and behavioral information which eluded us previously when using traditional clustering and regression-based approaches.

Neuronal information conductance often involves the transmission of action potentials. The spreading of action potentials along the axonal process of a neuron is based on three physical parameters: The axial resistance of the axon, the axonal insulation by glial membranes, and the positioning of voltage-gated ion channels. In vertebrates, myelin and channel clustering allow fast saltatory conductance. Here we show that in Drosophila melanogaster voltage-gated sodium and potassium channels, Para and Shal, co-localize and cluster in an area resembling the axon initial segment. The local enrichment of Para but not of Shal localization depends on the presence of peripheral wrapping glial cells. In larvae, relatively low levels of Para channels are needed to allow proper signal transduction and nerves are simply wrapped by glial cells. In adults, the concentration of Para increases and is prominently found at the axon initial segment of motor neurons. Concomitantly, these axon domains are covered by a mesh of glial processes forming a lacunar structure that possibly serves as an ion reservoir. Directly flanking this domain glial processes forming the lacunar area appear to collapse and closely apposed stacks of glial cell processes can be detected, resembling a myelin-like insulation. Thus, Drosophila development may reflect the evolution of myelin which forms in response to increased levels of clustered voltage-gated ion channels.