Dissemination of FlyWire, A Whole-Brain Connectomics Resource

This proposal will disseminate FlyWire, a Drosophila whole brain connectomics resource. We used advances in AI to segment all neurons from a whole brain EM volume called FAFB. The automated segmentation is of high enough quality that, in combination with innovative proofreading tools, scientists can relatively quickly proofread circuits of interest. The community of current collaborators includes about 160 scientists from 40 labs, who have so far succeeded at proofreading more than 15% of the neurons in the fly brain. Several publications have resulted, and more are on the way. Proofreading is required because automated segmentation, while good, contains errors that must be fixed with manual correction. Janelia/HHMI spent 50 person-years proofreading an EM volume that contains roughly one-third of the neurons in another female brain, a resource called the hemibrain. A whole brain connectome, in contrast, will make it possible to address questions about brain-wide circuit organization and about any brain region. We created this unique resource with a three-year grant from the BRAIN Initiative, and are now seeking additional funding to disseminate the resource more widely. Our framework for dissemination will set the standard for emerging, larger connectomics datasets, such as those of mammals. We propose three Aims for dissemination that include outreach efforts, deploying a centrally-managed proofreading team, and enhancing the user interface and tools available to access FlyWire data. In particular, a Community Manager will organize the community of scientists working in FlyWire, will recruit and train new users, will run workshops, and will coordinate access to the trained proofreaders. Scientists can apply to have an expert proofreader assist with their circuit reconstruction projects, and we will give priority to new users. Finally, in the first year of the project we will implement GUI tools in FlyWire that facilitate discovery without necessitating programming skills.