Utility of the Python package Geoweaver_cwl for improving workflow reusability: an illustration with...

Area of Science:

• Computational Science
• Artificial Intelligence
• Machine Learning

Background:

• Computational workflows are crucial for scientific data analysis, but reproducibility and reusability challenges hinder collaboration.
• Standardization is needed to address these issues, with the Common Workflow Language (CWL) emerging as a key framework.

Purpose of the Study:

• To develop a Python package that automatically converts AI/ML workflows from Geoweaver into CWL.
• To promote portability, reproducibility, and interoperability of AI/ML workflows.

Main Methods:

• Developed a Python package to translate Geoweaver workflows into CWL.
• Tested the package on diverse use cases across different scientific domains.
• Made all code and datasets publicly available.

Main Results:

• The Python package successfully describes AI/ML workflows in CWL format.
• Demonstrated the utility and effectiveness of the package through multiple use case validations.
• Confirmed the package's ability to facilitate well-versed AI processes.

Conclusions:

• The developed Python package enhances the standardization and reusability of AI/ML computational workflows.
• The package promotes better collaboration and data sharing within the scientific community.
• Identified opportunities for future extensions and improvements to the workflow conversion process.