CWL-Bubble: Extending CWL for Dynamic Scientific Workflows

With the rapid growth of data size and computational complexity in modern scientific research, automating and ensuring the reproducibility of workflows has become crucial. Workflow management systems (WMS) are essential for efficiently managing large-scale workflows across domains like bioinformatics, geology, and astronomy. However, the proliferation of domain-specific WMS has created integration challenges, often leading to significant overhead. The Common Workflow Language (CWL) standard addresses this by providing a platform-independent framework for workflow portability and reproducibility. Despite its adoption, CWL struggles with supporting dynamic constructs, such as loops and recursion, which are essential for iterative and data-intensive workflows. This paper proposes CWL-Bubble, an extension mechanism to CWL that introduces a dynamic node, “Bubble”, that allows runtime generation of new tasks by external scripts. This approach enhances workflow adaptability and supports advanced control-flow structures such as iterative refinement and task decomposition. We demonstrate the effectiveness of CWL-Bubble in scientific case studies, showing that its dynamic features improve execution time in complex workflows while maintaining compatibility with existing CWL standards.