Pan-neuro: interactive computing at scale with BRAIN datasets

New technical and scientific breakthroughs are enabling neuroscientific measurements that are both wider in scope and denser in their sampling, providing views of the brain that have not been possible before. At the same time, funding initiatives, as well as scientific institutions and communities are promoting sharing of neuroscientific data. These factors are creating a deluge of neuroscience data that promises to provide new and meaningful insights into brain function. However, the size, complexity, and identifiability of the data also present challenges that arise from the difficulties in storing, accessing, processing, analyzing, visualizing and understanding data at large scale. Based on their successful adoption in the earth sciences, we have started adopting and adapting a set of tools for interactive scalable computing in neuroscience. We are building an approach that is based on a combination of a vibrant ecosystem of open-source software libraries and standards, coupled with the massive computational power of the public cloud, and served through interactive browser-based Jupyter interfaces. Together, these could provide uniform universal access to datasets for flexible and scalable exploration and analysis. We present a few prototype use-cases of this approach. We identify barriers and technical challenges that still need to be addressed to facilitate wider deployment of this approach and full exploitation of its advantages.