MIRAPIE Guideline

MIRAPIE Guideline Required tasks and supportive information for setting up and automating harmonised provenance documentation in biomedicine Authors: The MIRAPIE community Interactive implementation: Lea Michaelis¹, Benjamin Winter¹, Judith AH Wodke¹ ¹ Medical Informatics Laboratory, University Medicine Greifswald Version: 2025.0 Introduction Provenance information documents the transformation history for a data object of interest along its data life cycle. In line with the FAIR principles [1], provenance information is providing a transparent basis for the assessment of data object quality and (re)usability. According to our aim of broad applicability across research domains, a data object can be a single measurement, a data set, a software tool or processing pipeline, as well as a bioprobe, an experimental protocol or any other "named thing" ([2]) that is processed and used (in biomedicine). The  MIRAPIE  (MInimal Requirements for Automated Proven‐ ance Information Enrichment) project aims to harmonise and foster provenance information documentation in biomedicine [3]. The MIRAPIE question "WHO did WHAT using WHICH tools HOW, WHY, and WHEN with data objects obtained from WHERE?" summarises seven purpose-driven questions derived from previous publications [4, 5, 6] thus providing a harmonising framework for mapping any domain-specific provenance inform‐ ation to the seven W-keywords. Provenance information should be documented according to a domain-specific data model cap‐ able of describing required domain-specific features of the docu‐ mented data processing. For the biomedical fields, the MIRAPIE project provides a minimal consensus data model and an accompanying upper-level ontology [7]. The MIRAPIE model can be easily extended by required domain-specific sub-classes. Al‐ ternatively, any other more suitable data model, such as the REPRODUCE-ME ontology [8] or the Common Provenance Mod‐ el [9] can be mapped. In brief The MIRAPIE guideline, while aiming to be as widely applicable as possible, supports easy entrance to (automated) provenance documentation specification of provenance information required at different stages of the data life cycle automation of provenance information documentation harmonisation of provenance information based on the MIRAPIE question interoperability of provenance information across biomed‐ ical disciplines and national borders • • • • • MIRAPIE Guideline © 2025 by MIRAPIE community is licensed under CCBY4.0. To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0 https://codeberg.org/MIRAPIE/MIRAPIE 1/4 MIRAPIE Guideline While within the MIRAPIE project the focus lies on provenance information for biomedical data objects, the here presented guideline is generally applicable, as long as the documented provenance information can be mapped to the research fieldbridging MIRAPIE question. High-quality provenance information is specified in a team effort, bringing together the expertise of researchers, data managers, and data stewards. The MIRAPIE guideline provides a checklist of actionable steps for i)specifying, ii)modelling, iii)automating, and iv)harmonising provenance information documentation. To optimise such documentation and to facilitate the assessment of the provenance maturity reached, these action steps are organ‐ ised according to established maturity levels [10]. Each step is complemented with additional information and practical tips, for example about the required expertise, specific aspects to con‐ sider for local implementation, or adherence to the FAIR prin‐ ciples [1]. The MIRAPIE guideline has been designed to ease entrance into the provenance topic on this purpose-driven basis and to estim‐ ate the provenance maturity reached within a project, for a pro‐ cessing pipeline, or institution-wide. This guideline does not provide a generally applicable solution for the local technical im‐ plementation of automated provenance information document‐ ation nor for the quality assurance of collected provenance in‐ formation. Checklist Maturity level 0 -- unmanaged Being aware of the relevance of provenance information for data quality and data interpretation in context without having started to actively document provenance information. (This level is already reached when starting to set up provenance documentation.) Become aware of the relevance of provenance information for data quality and data interpretation in context. Maturity level 1 -- incipient This level is reached by specifying the relevant provenance information for a data object of interest within the research context. Expected output: informal document. Specify the minimal information required for describing the history of a data object of interest along its data life cycle. Define a level of detail for the anticipated (re)usage scenarios or desired provenance information quality, e.g., documentation of every processing step applied to the data object along its life cycle. Consider different perspectives, such as clinical decision making and reuse in research for a routine care measurement. This task should be accomplished involving expertise about the biomedical domain and the data processing, e.g., the researcher or research team with support of a data steward and a data manager. • • • 2/4 ✔ Maturity level 2 -- controlled Level 2 is reached when provenance characteristics have been established by answering the MIRAPIE question for your data object of interest. Expected output: visualisation of the data object history, improved data management plan. Answer the MIRAPIE question at least once for your data object of interest and possibly for every transformation step of its data life cycle. Map the specified information carefully to a data model of choice, e.g., the minimal MIRAPIE model. Structure the information according to the W-question information categories, e.g., “WHO”: “University Medicine Greifswald, Core Unit Data Integration Center”. Draw a respective sketch to visualise the provenance information. We recommend to follow an established visualisation schema, e.g., the PROV-DM, using any tool that supports generation of vector graphics(eps, svg). Please follow the standard notation for provenance information, for example by following the Provenance Notation  (PROV-N) from W3C[11]. Ideally, the sketch is generated in an iterative process: Start from a summary perspective, e.g., transforming and enriching clinical data for reuse in research, and extend this perspective into depth by subtasks until reaching the predefined level of detail,  e.g., each processing step applied to the data object along its life cycle. Document this specified information, e.g., in your data management plan. Open data initiatives as the Research Data Alliance support data management plan generation [12]. Maturity level 3 -- operational This level requires the implementation of provenance information documentation into a respective annotation or processing pipeline. Expected output: provenance graph in an appropriate structured format. Integrate provenance tracking technically into your data processing pipeline to enable automated generation of the desired provenance information. Accomplishing this step does depend on the data object type and on the local infrastructure for data processing. For a manual laboratory measurement, the measuring person will have to manually document related provenance information. Using structured and possibly pre-filled forms minimises manual effort and error probabilities, annotation tools to automatically enrich data and metadata can further improve their quality. The most suitable tools and forms will likely be domain-specific. For a digital data processing pipeline that automates data collection and processing, related provenance information tracking can be directly integrated into the pipeline. This has been prototypically tested for example at German data integration centers [13, 14]. When provenance documentation has been automated previously with using a different data model, maturity level&nbso;3 is MIRAPIEindependent already reached. In this case, harmonisation is achieved by adding an automatic mapping feature to the existing processing pipeline that maps the automatically collected provenance information to the MIRAPIE question according to your specification for maturity level&nbso;2. Deliver at least some kind of centralised(cross-tool) collection of provenance information in an interoperable format that is reliably long-term-stored and can be queried independently of individual tools. The domain-specific provenance information should be provided according to FAIR principles [1] and community standards, e.g., as RDF specification. The mapping result should be stored in a structured format, for example using a YAML or JSON file. For examples, please check out our usage scenarios on codeberg. • • • • • • • • • • • • 3/4 Maturity level 4 -- optimised This level requires that provenance information is automatically processed and made available for reuse according to applicable community standards and to the FAIR principles [1]. Write scripts to automatically parse and process the documented provenance information and to transform data and metadata into formats compatible within your research context. Ideally, this step is achieved in collaboration with different domain experts, including data collectors, data managers, and data users. In medical context, the findability and accessibility of data and associated metadata might be strictly regulated and restricted to a defined group of authorised people. Still, within the legal restrictions and given the extra effort required to obtain the data, provenanceimproved data objects are particularly desirable. Evaluate your tracked provenance information. To confirm content-specific utility of the collected information, share it with an expert from a different domain also working with the data of interest, but possibly in a different context. Please provide this person the MIRAPIE question and ask for feedback concerning understandability and plausibility of the documented provenance information. Would the collected information provide your colleague enough context for assessing basic data quality metrics and evaluating the data objects utility? Desirable: Publish your provenance information. Please adhere to the FAIR principles and whenever permitted provide data object and metadata, including provenance information, for long-term reuse. A license that permits data object reuse is essentially required for others to reuse the data object. Creative Commons provides support with choosing your license. A persistent identifier and publication in research databases or public data repositories provides the basis for long-term accessibility of the data. For example, Zenodo.org allows to publish from datasets to guidelines. References Wilkinson MD, et al. The FAIR Guiding Principles for scientific data management and stewardship. Scientific data 2016; 3. Unni DR, et al.; Biomedical Data Translator Consortium. Biolink Model: A universal schema for knowledge graphs in clinical, biomedical, and translational science. Clin Transl Sci. 2022 Aug;15(8):1848-1855.DOI: 10.1111/cts.13302. Gierend K, et al. TAPP: Defining standard provenance information for clinical research data and workflows - Obstacles and opportunities. Companion Proceedings of the ACM Web Conference 2023. https://api.semanticscholar.org/ CorpusID:258377678. Ram S, Liu J. Understanding the Semantics of Data Provenance to Support Active Conceptual Modeling. In: Chen PP, Wong LY, eds. Active Conceptual Modeling of Learning. Berlin, Heidelberg: Springer Berlin Heidelberg, 2007: 17–29. Buneman P, et al. Why and where: A characterization of data provenance. In: Database Theory-ICDT 2001: 8th International Conference London, UK, January 4– 6, 2001Proceedings 8. Springer, 2001: 316–30. Herschel M, et al. A survey on provenance: What for? What form? What from? The VLDB Journal 2017; 26: 881–906. Genehr S, et al. MIRAPIE Ontology - MInimal Requirements for Automated Provenance Information Enrichment in biomedical research. 2025; published online May 16. DOI:10.5281/ZENODO.15608384. Samuel S, König-Ries B. REPRODUCE-ME: Ontology-Based Data Access for Reproducibility of Microscopy Experiments. In: Blomqvist E, et al.  (eds) The Semantic Web: ESWC 2017 Satellite Events. Lecture Notes in Computer Science, vol 10577. Springer, Cham. DOI:10.1007/978-3-319-70407-4_4; IRI: https://w3id.org/reproduceme. Frexia F, et al. The common provenance model: Capturing distributed provenance in life sciences processes. InChallenges of Trustable AI and Added-Value on Health 2022 (pp. 415-416). IOS Press. DOI:10.3233/shti220489. Gierend K, et al. The Status of Data Management Practices Across German Medical Data Integration Centers: Mixed Methods Study. J Med Internet Res 2023; 25: e48809. https://www.w3.org/TR/prov-n https://www.rd-alliance.org Gierend K, et al. Traceable Research Data Sharing in a German Medical Data Integration Center With FAIR (Findability, Accessibility, Interoperability, and Reusability)-Geared Provenance Implementation: Proof-of-Concept Study. JMIR Form Res. 2023 Dec7;7:e50027. DOI:10.2196/50027. Johns M, et al. Tracking provenance in clinical data warehouses for quality management. International Journal of Medical Informatics 2025, 193, 105690. DOI:10.1016/j.ijmedinf.2024.105690. • • • • • • • 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 4/4