Connectivity Is All You Need: Inferring Neuronal Types with NTAC

Abstract

Recent advances in electron microscopy and computer vision have enabled the mapping of complete wiring diagrams--called connectomes--of brain regions and even entire brains. The emergence of these increasingly large-scale connectomic datasets have intensified the need for efficient and accurate neuronal cell type identification. Traditional approaches rely on labor-intensive analyses of molecular, anatomical, and physiological features. As a step toward fully automated neuronal cell type classification, we present NTAC (Neuronal Type Assignment from Connectivity) -- a method for grouping neurons into cell types based solely on synaptic connectivity. Our approach is grounded in the hypothesis that synaptic connectivity is key to determining neuronal cell types (Seung 2012), and our results provide the strongest evidence to date supporting its validity. NTAC comes in two flavors: a semi-supervised variant that requires some of the neurons to be labeled, and an unsupervised variant that requires no labels at all. The first can be naturally formalized as a learning problem on graphs: given labels for a subset of neurons, use connectivity to label the rest. We present a simple and fast algorithm that achieves over 95% accuracy on the flys visual system starting with just 2% of neurons labeled, within minutes on a standard PC. In contrast, morphology based parallel achieves lower accuracy even on smaller datasets and with significantly more labels. Formalizing the problem as a fully unsupervised learning task is more challenging. To address this, we introduce a novel computational problem, approximate equitable partitioning, and design an efficient heuristic for it, using the semi-supervised algorithm as a subroutine. Remarkably, the unsupervised method achieves [~]70% accuracy on the flys visual system, drastically outperforming its morphology based parallel and providing concrete evidence that synaptic connectivity alone can reveal cell types. Both variants of NTAC are evaluated on multiple state of the art connectomes including optic lobes, central brain and nerve cord of adult fruit flies.