This deliverable demonstrates how the EPND Hub can support the preparation of data and metadata for sharing, discovery and analysis. Using data from EPND Case Studies 1, 2 and 4, the report investigates which stages of the Extract, Transform, Load (ETL) and harmonisation process can be supported by software tools, and where specialist input remains necessary.

The case study exercise brought together data from nine cohorts contributed by four consortium partners and two external partners, with approximately 120 data items requested from each cohort. Differences in file structures, coding, units and documentation required cohort-specific processing. The report distinguishes three main stages: gathering data, transforming their structure and values, and mapping data elements to standard terminology.

Three mapping tools were examined: FAIRnotator, developed within EPND, the NTK-Curation App, and a harmonisation tool previously available on the AD Workbench. All three were successfully deployed in AD Workbench testing workspaces. Their different capabilities supported an approach in which researchers can select and deploy tools suited to their study, rather than being required to use a single harmonisation solution.

The report also assesses the Observational Medical Outcomes Partnership (OMOP) Common Data Model using eight variables from the previously harmonised case study dataset. This demonstrated technical feasibility for the selected variables, while identifying limitations in the availability of standard vocabulary terms for neurodegenerative disease research.

A further exercise explored metadata syndication between the AD Workbench and EPND Catalogue. Differences in required fields and their meanings necessitated manual mapping and an adapted presentation of syndicated entries.

The report concludes that the Hub should facilitate harmonisation through flexible tool deployment and guidance. It also identifies metadata harmonisation as a practical priority for expanding catalogue coverage without requiring contributors to transform their underlying datasets into a single common data model.

For more information, find the full Deliverable report here.