In this presentation, I will review some of the challenges that neuroscience and
neuroimaging are facing, and in particular the challenges of reproducible and generalizable studies in the context of data distributed and under variable legal and ethical constraints. The field of neuroscience is facing new questions on the impact of variable processings and the difficulty to generalize results. These challenges have both technical and sociological roots, and the solutions will therefore likely also be both technological and sociological. I will take some examples of neuroinformatics projects developed in my laboratory or in collaboration to transform how we share and analyze data with an example of the development of imaging biomarkers for Parkinson Disease. I will show preliminary results showing that distributed infrastructures can be leveraged for federated learning of Parkinson Disease progression. I will open the discussion on some key aspects to consider, for instance how these infrastructures can be maintained, how their governance structure can be thought of, and how this offers opportunities to build community of practices.