Registries of domain-relevant semantic reference models help bootstrap interoperability in domains with fragmented data resources

The specialist field of rare diseases must connect its vast array of globally distributed disease and patient registries to maximise their value. Unfortunately, many registries are "boutique", with few or no staff with formal informatics training. At a series of Bring Your Own Data workshops, we helped registry owners transform their data into formally structured triple stores following the Linked Data principles and demonstrated the potential of data linkage. We documented several useful approaches that we believe could be followed independently by other registry owners worldwide, including: that the transformation to Linked Data could be considered as passing through layers of increasing semantic complexity; that only a subset of ontologies are relevant at each layer; and that certain data transformation processes could be modelled as an "archetype", and presented to registry staff to fill-in with their data. We propose that formally capturing these ontological layers and archetypes, and registering them as a reference and teaching resource will facilitate the wider community of non-expert data owners self-direct their own data transformations.