A Non-Intimidating Approach to Workflow Reproducibility in Bioinformatics: Adding Metadata to Research Objects through the Design and Evaluation of Use-Focused Extensions to CWLProv

Thesis and associated data and code for my master project. Abstract In the era of big data and big science, workflows have been proposed as a means to achieve computational reproducibility. However, sharing the results of a workflow execution is challenging. Research Objects (ROs) can solve this problem by packaging the workflow description and the data associated with an analysis, together with structured metadata describing the meaning and relationship between the contained entities. ROs contain the provenance of the generated results. However, there is a need for a standard of metadata which should be contained in ROs. Here, we define a taxonomy of provenance types based on a use case Bioinformatics workflow. Subsequently, we assess CWLProv 0.6.0, a standard for sharing the execution results of CWL workflows as a RO, for the representation of each of the components of the taxonomy. Based on the results of this analysis, we propose a standard for the annotation of input data, as well as an extension of the provenance graph to enable richer annotations. We are confident that the work described here is not only relevant for CWLProv but can also be applied to other RO specifications (such as the Workflow Run RO-Crate profile). A standardized approach to the representation of workflow executions can serve as the basis for automated extraction of the collected provenance. Ultimately, this is one step closer to achieving computational reproducibility. GitHub repositories: https://github.com/RenskeW/cwl-epitope https://github.com/RenskeW/cwlprov-provenanc