This talk was presented at the 6th Joint Congress of the DGBP and AGNP, held in Berlin from 6 to 8 May 2026 under the theme “Neuroscience for Mental Health.” It followed an introductory lecture by Ute Habel on the research plan for aggression in mental disorders and the scientific ambitions of TRR379.
The talk presented the information-management concept developed within Q2 of TRR379 as a service-oriented infrastructure for a large, multi-site translational research network. The central aim was to create a practical, sustainable system that supports research across six sites, around 1,000 individuals volunteering for this research, and a longitudinal study horizon extending over more than a decade. Rather than imposing a rigid centralized workflow, the approach treats the consortium as a network of mutually responsible actors and seeks to maximize scientific benefit while minimizing the burden of contribution. The resulting framework emphasizes independently governed, adoptable solutions that can fit different local conditions without losing interoperability.
A major focus of the talk was the challenge of heterogeneity in a distributed research center. Site-specific implementations, differing procedures, and changing equipment, personnel, and legal contexts make standardization difficult. Q2 addresses this by building a self-hostable service stack for project and data hosting that can accommodate local autonomy while still enabling shared structures. The tools introduced in this context included Forgejo-anaksajo for version-control, and DataLad for client-side provenance tracking. Together, these components provide a foundation for managing research data, code, and metadata across sites in a way that remains traceable and adaptable.
The talk also emphasized the role of metadata as the backbone of collaborative research management. The metadata system supports both manual browser-based entry and programmatic submission, covering not only research data but also administrative information and outreach-related records. Its design combines machine-validatable structural models with manual curation by authorized experts, allowing automated checking where possible while preserving domain-specific oversight where needed. Additional functions such as profile image uploads and publication registration via DOI, matched against membership records and ORCIDs, illustrate how the system links scientific outputs, people, and organizational context.
A key conceptual contribution was the idea of a central collaboration hub that allows browser-based navigation across datasets and research stages. This makes it possible to trace data generated at one site back to original acquisitions at another, supporting full provenance from acquisition to publication. The talk described how derivative data produced in Jülich can be traced to original acquisitions in Aachen, Mannheim, Heidelberg, and Frankfurt, and how the broader data graph links initial DICOM files to final analyses reported in papers. The system records not only what was computed, but also how it was computed and in which computational environment, thereby enabling independent verification and robust reproducibility.
Another important theme was quality control. The infrastructure supports both manual and programmatic QC, including automated triggers and custom visualizations such as brain exploration tools. This combination of human expertise and machine-assisted checking was presented as essential for maintaining quality in complex, long-running studies. The approach is meant to make collaboration easier while still ensuring that data handling remains scientifically rigorous and transparent.
The talk further situated Q2 within a broader ecosystem of open-source tools and collaborative infrastructures. Its toolkit originated in the ABCD-J project and is intended to be usable beyond the immediate TRR379 context, including in brain sciences, biology, or climate science. Looking ahead, the system is designed not only for internal coordination but also for external discovery and secure computation. Planned developments include exporting information into discovery databases such as NeuroBagel to connect with interoperable sites and support artificial cohort generation, as well as enabling external analysis scripts to be run locally at the data site so that sensitive clinical data never needs to leave its origin.