Singularity Containers Improve Reproducibility and Ease of Use in Computational Image Analysis Workflows

Area of Science:

• Computational Biology
• Bioinformatics
• Microscopy Image Analysis

Background:

• Reproducing computational workflows in scientific image analysis is challenging due to software versioning and dependency issues.
• Users with limited scientific computing knowledge face significant hurdles in replicating complex image analysis pipelines.
• The adoption of containerization technologies like Singularity in image analysis remains limited.

Purpose of the Study:

• To introduce Singularity containers as a robust solution for reproducible and shareable image analysis workflows.
• To provide a practical, step-by-step protocol for packaging a segmentation algorithm into a container.
• To facilitate the use of advanced image analysis tools on high-performance computing (HPC) resources.

Main Methods:

• Developed a detailed protocol for packaging a state-of-the-art image segmentation algorithm.
• Utilized Singularity container technology to encapsulate the algorithm and its dependencies.
• Demonstrated the process on a local Windows machine for deployment on an HPC cluster.

Main Results:

• Successfully packaged a complex image analysis workflow into a portable Singularity container.
• The containerized workflow ensures consistent execution across different computing environments.
• The protocol enables users to easily run advanced segmentation tools on HPC infrastructure.

Conclusions:

• Singularity containers offer a powerful and accessible method for ensuring reproducibility in image analysis.
• This protocol lowers the barrier to entry for utilizing sophisticated computational tools in microscopy research.
• Increased adoption of containerization can significantly advance collaborative and reproducible scientific discovery in image analysis.