Project Summary Anatomical atlases are spatial reference maps of cells in tissues/organs/brains and provide structure information for a wide range of biological analyses. The anatomical atlas of C. elegans nervous system is the only atlas for the entire nervous system of an animal with a resolution of all neuronal classes. However, built on a limited dataset and manual annotations, the standard atlas is insufficient in capturing biological variabilities, inaccurate and difficult to use for cell identification routinely, and only applicable for wildtype adult. While several heroic efforts of generating and imaging marker strains to build atlases have much improved the atlases, there is still a need for a pipeline to build accurate genetic-background-specific (or experimental-condition-specific) atlases easily and cheaply; further, there is a need to build such atlases that can be used without specialized equipment and with as few genetic perturbations as possible. Recent development of machine learning techniques and molecular transgenic approaches enabling the systematic production of in vivo reporters and imaging methods capable of collecting and processing high-resolution datasets at a large scale. The goal of this application is to address the current bottleneck by establishing a combined experimental and computational pipeline for modularly built, complete, coordinate- and template-free brain atlases for democratized and flexible uses. By imaging in vivo markers in a large number of live animals, the project will generate complete anatomical atlases for the C. elegans nervous system that capture variability in the population, which will greatly enhance the accuracy of identity predictions when used on each animal. The project will generate a collection of transgenic animals expressing partly overlapping in vivo markers that cover all neurons and build a computational pipeline to assemble the atlases. Further, a few widely applicable developmental atlases as a direct output of the project will showcase the pipeline and the approach. Importantly, the atlases do not seek to provide a set of rigid coordinates for each neuron class, but instead, a set of constraints that can be used to provide best estimates of neuron identities for each new sample. This ensures accuracy and applicability of the atlases to specific use case. The building of whole-brain atlases is piece-wise from easily-obtained partial atlases, and can be crowd-sourced if desired. The use will be streamlined with image input and neuron-identity prediction output. The proposed project is innovative, because it will build the first complete anatomical atlases of a nervous system using large datasets collected from in vivo markers of many live animals; it uses relational information uniquely suited to provide more accurate assignments; it will capture variabilities among individual animals. The proposed the work is significant, because it will address the urgent and unmet need for a...