Understanding how the brain generates behavior requires two types of very sharp lenses: one focused on neural activity, and the other on the animal itself as it interacts with the world. The lenses available to us for peering at the brain and for characterizing behavior have steadily improved, which is both a boon and a curse: a boon because it is likely that key insights will arise from neural and behavioral data that approximate each other in terms of resolution, but a curse because organizing and modeling such dense and diverse data is a significant challenge. The Projects in this grant combine advanced techniques in machine vision, machine learning and neural recordings to address how neural activity in midbrain and striatal circuits orchestrate the behavior of rodents and control its evolution. The experimental aspects of this project are to be carried out by researchers located both in the Boston and the Bay area. Here we propose a Data Science Core (DSC) that will enable these diverse research groups to store, retrieve, and analyze their data using modular programs that can be hot- swapped for easy experimentation and iteration. In this structural role the DSC will adhere to a series of core principles for open science, including the use of open source code and data standards whenever possible, creating frameworks that allow for easy data sharing, and the careful curation of metadata to allow pooling of information from multiple labs to enable large-scale statistical testing and validation. The DSC will also be responsible for the distribution of generated data (including raw data, metadata, and analyzed datastreams) to the broader community, including to other U19 consortia. The DSC will therefore sit at the center of this proposal, acting as a crucial hub through which data and insight pass freely between investigators and labs.
