Neural cell-type identification is crucial for understanding brain function and developing targeted therapeutics for brain dysfunction. While modern silicon probes can record hundreds of neurons simultaneously across all cell-types indiscriminately, they provide limited information about neuron identity. I will discuss how, through a multi-lab collaboration, we developed a strategy to reliably predict cell-type identity of cerebellar cortical neurons from their electrophysiological signatures. We used optogenetics to establish a ground-truth cell-type database and deep learning to build a semi-supervised classifier that can generalize across ​regions and species.

This event is part of the BCM Data Science Community monthly seminar series.