SIMPROV : Provenance capturing for simulation studies

Improving interpretability and reusability has become paramount for modeling and simulation studies. Provenance, which encompasses information about the entities, activities, and agents involved in producing a model, experiment, or data, is pivotal in achieving this goal. However, capturing provenance in simulation studies presents a tremendous challenge due to the diverse software systems employed by modelers and the various entities and activities to be considered. Existing methods only automatically capture partial provenance from individual software systems, leaving gaps in the overall story of a simulation study. To address this limitation, we introduce a lightweight method that can record the provenance of complete simulation studies by monitoring the modeler in their familiar yet heterogeneous work environment, posing as few restrictions as possible. The approach emphasizes a clear separation of concerns between provenance capturers, which collect data from the diverse software systems used, and a provenance builder, which assembles this information into a coherent provenance graph. Furthermore, we provide a web interface that enables modelers to enhance and explore their provenance graphs. We showcase the practicality of SIMPROV through two cell biological case studies. Author summaryWith the importance of simulation studies in understanding and managing complex dynamic systems, the need to support the interpretation and (re-)use of their results increases. Provenance documents how the products of a simulation study were created and what other products, agents, and activities have been involved in this process. For example, the information based on which data from which cell line a simulation model has been calibrated and validated is central to interpreting the results and assessing how the results can be reused. Therefore, some software tools offer to record provenance information. However, for complete provenance information, the tool must offer all functionalities required for a simulation study. In practice, various tools are typically used. To accommodate this situation, we propose a flexible, decentralized approach: SIMPROV. A provenance capturer - a small piece of software designed to record the modelers actions within a software tool - observes each tool used by the modeler. A central provenance builder then combines the recorded information from all captures. A capturer has to be programmed only once for each software tool used in systems biology, and modelers can work as before with minimal effort needed to record the provenance of their simulation studies automatically.