AVOCADO: Visualization of Workflow-Derived Data Provenance for Reproducible Biomedical Research

Area of Science:

• Biomedical Informatics
• Data Visualization
• Computational Biology

Background:

• Data provenance is crucial for reproducibility in data-driven biomedical research.
• Existing node-link diagrams for provenance visualization do not scale to complex multi-step analysis workflows.
• Visualizing large provenance graphs is challenging for hundreds of samples and numerous workflows.

Purpose of the Study:

• To develop a scalable and effective method for visualizing complex data provenance graphs in biomedical research.
• To enable users to review and communicate multi-step analysis workflows.
• To address the limitations of current node-link diagram approaches.

Main Methods:

• Implemented hierarchical and motif-based graph aggregation to reduce complexity.
• Developed a modular degree-of-interest (DoI) function for adaptive graph expansion based on user interaction.
• Integrated the approach into an analysis platform capturing extensive data provenance information.

Main Results:

• The proposed interest-driven adaptive visualization effectively handles complex provenance graphs.
• Hierarchical and motif-based aggregation reduces visual complexity.
• The system allows users to explore relevant parts of the provenance graph efficiently.

Conclusions:

• The developed provenance visualization approach enhances the review and communication of complex biomedical analyses.
• This method improves reproducibility by providing a scalable way to represent data provenance.
• The interest-driven visualization is effective for data-driven biomedical research scenarios.