Establishing a research data management system at BAM: Challenges and approaches

The Bundesanstalt für Materialforschung und -prüfung (BAM, federal institute for materials science and testing) contributes to the standardization efforts of research data in various domains of materials science and engineering. Since 2021, a research data management (RDM) system (BAM Data Store) is being established to enable central storage of (meta)data generated during the active phase of the data lifecycle and enable compliance with the FAIR principles[1]. Since May 2023, after a successful pilot phase, the departments of eScience and IT infrastructure started the institute-wide rollout with the aim of integrating newly generated research data from up to sixty scientific units over a period of three years. In this abstract, we review the components required to develop an institutional RDM service. We list challenges and lessons learned over a year to prioritize tasks and identify resources needed for data stewardship to achieve adoption of the RDM system. The BAM Data Store: an institutional RDM system based on openBIS software At BAM, research, development, and technical services are conducted with high societal impact by a scientific staff of more than 1,000 employees. The BAM Data Store is based on openBIS[2], [3], an open-source RDM software developed at ETH Zurich (ETHZ) that includes plugins for the digital representation of laboratory inventory, and an ELN. The openBIS framework, allows for the storage of meta(data), and provides interfaces for the integration of data analysis. At BAM, three to four divisions are enrolled every four months. Data store stewards (DS) and users are trained to use the software and develop first implementations. Under the guidance of the DS and the division lead, all users are responsible for implementing the data store to store newly generated research data according to RDM best practices to generate self-describing and traceable data sets. Development of RDM Services For the development of RDM services, eight components were previously listed[4]. These are relevant before (1), during (2) or after (3) data collection (Fig 1). With the aim of listing challenges identified over the course of a year and sharing the approaches to establishing an institutional RDM system at BAM, we address the components relevant during data collection: “Managing active data” and “Guidance, Training, and Support”. Challenges and approaches of developing RDM services During one-year of training users in the RDM system (BAM Data Store), we identified challenges faced to develop the RDM service for supporting management of active data. The BAM Data Store is primarly set up to document and store data in the “Collection” phase of the Data life cycle. To manage data during “Processing” and “Analysis”, the system allows the integration of Jupyter Notebooks. As the system is set up to preserve data with access restricted to institutional members, data preservation in external data repositories is encouraged. Table 1 lists the challenges posed for developing the BAM Data Store as an institutional RDM service. The current focus is on set the challenges that arise during data collection. 1. Managing active data Active Data Management Challenges Approaches Collect Automatic data migration from measuring devices to the RDM system 1. Development of underlying software function (openBIS-Dropbox) 2. Support in writing scripts for data migration Process Standard documentation of data processing 4. Support and train the use of tools (Jupyter Notebooks and openRefine) Analyze Documentation of data analysis 5. Support and train the use of Jupyter Notebooks Preserve Standard preservation 6. Define workflows, support and train for the use of Gitlab and the preparation of data for export to external data repositories Table 1. Challenges and Approaches of Establishing RDM services: Active data management covers all steps of the data lifecycle before the data is published. Four challenges and approaches for prioritising resources for training and support are listed. 2. Guidance, Training and Support The training and support concept was developed in 2023 to enable researchers to become Domain Data Stewards (DDS) with the aim of introducing the institutional groups to a centralised RDM system while ensuring adoption of the tool 5. The onboarding concept was tested for one year with seven groups identifying different levels of implementation in each group. We identified two main challengues for the implementation of the RDM system and thus for its development as a service: 1. the time needed by the DDS to customise the system, define and coordinate the RDM processes within the team. 2. the different levels of knowledge about concepts for the digital representation of data and data modeling. Although these concepts are more intuitive for researchers with strong programming background, communication between colleagues in multi-disciplinary contexts remains a challenge. To overcome these challenges, the concept was further developed before a new onboarding phase began in 2024. The challenges and approaches in terms of "guidance, training and support" are listed in Table 2 for four strategic steps for the implementation of the RDM system based on the openBIS software. We have started with an overarching strategy to guide data modeling workflow and improve implementation. For this, we introduced a new training module for “Data Model” and defined a successful onboarding phase as the one in which the DDS implements a use case. We also motivate this work by making it visible through the publication of data models and use cases. This process is supported by defining a publishing workflow, setting up an institutional GitLab repository and providing guidelines to make these datasets citable and reusable. Strategic steps Challenges Approaches Data Model Mapping research data and related workflows in RDM software 7. Add training module – Data Model. Includes a data model guideline and supporting videos with use cases Masterdata Identify elements of the research processes that should be represented as entities and coordinate teamwork for iterative metadata development 8. Provide a collaborative tool to guide the team in identifying masterdata elements. RDM system customization Strongly dependent on the previous steps. Correlation of data management needs and software functions 9. Provide webinars with hands-on session to train software functions 10. Motivate implementation by definining and communicating the output of a sucessful onboarging: use case implementation Implementation Strongly dependent of Masterdata and system customization 11. Define a workflow to support the publication of data models and use cases developed during and after onboarding 12. Plan resources for post-onboarding implementation support Table 2. Challenges and Approaches of Establishing RDM services: Guidance, training and support are the foundation for setting up an RDM service once the infrastructure is established. Challenges and approaches are listed for four strategic steps towards implementation thus establishment of a RDM service. The second version of the training and support concept contains five modules to train researchers to become Domain Data Stewards (DDS) with the aim to develop the BAM Data Store in an institutional RDM system. The modules are listed to give an overview of the objectives and changes of the concept compared to the first version described in 2023[5]. 1. Communication module: Remains as initially defined. It aims to identify the RDM needs of each individual research group and to keep stakeholders informed throughout the process. 2. Data model module: It is new. It aims to train DDS with any background knowledge to digitally represent research processes by mapping them in the openBIS-system. In addition to support DDS in coordinating this work with their colleagues. 3. Masterdata module: Contents have been divided into Masterdata and Data model -modules. The Masterdata module aims to train researchers with any background knowledge of metadata concepts, with a focus on the definition of openBIS Masterdata. 4. Implementation module: It was previously called “hands-on module”. The module aims to train in hands-on sessions specific openBIS-software functions required to enable DDS to customize Inventory and ELN for the group. 5. Community-building module: The aims have been newly defined. The module aims to make visible the implementation of the RDM system by the researchers during and after onboarding. Data Stewardship and RDM tool adoption Acceptance of tools depends largely on how relevant the tool is to the users, whether it can fulfil the users' expectations, how user-friendly the tool is, and how well users are guided, trained, and supported. Data stewards face challenges related to these aspects, which should be carefully considered when making decisions about RDM tools, when establishing on communication strategies with institutional leadership and users, and when developing guidance, training, and support programs. After a year of onboarding BAM users into the central RDM system, several challenges have been identified that hinder tool adoption and implementation. Supporting data management during data processing and analysis requires digital representation of the collected data and lies at the interface between agnostic and discipline-specific data management, as it depeds on how the researchers analyse data and what skills they have. Therefore, we currently prioritize challenges and approaches identified for “Guidance, training and tupport” with a focus on RDM needs during data collection and preservation which can be addressed with the RDM system. Additional resources need t