Managing genomic variant calling workflows with Swift/T.

Area of Science:

• Bioinformatics and Computational Biology
• Genomics and Personalized Medicine

Background:

• Bioinformatics research relies on complex, multi-step workflows.
• Managing these workflows on computer clusters for large-scale data is challenging.
• Existing workflow management systems face limitations in scalability and portability.

Purpose of the Study:

• To evaluate Swift/T as a workflow management system for bioinformatics.
• To implement and assess a genomic variant calling workflow using Swift/T.
• To identify design criteria for robust, scalable, and maintainable production-grade workflows.

Main Methods:

• Implemented a primary genomic variant calling workflow in Swift/T.
• Focused on workflow management, performance, and scalability for big data.
• Introduced novel features to the Swift/T language.
• Formalized design criteria for production-level workflows.

Main Results:

• Swift/T demonstrated transparent operation across multiple cluster environments (PBS Torque, SLURM, etc.), ensuring workflow portability.
• The system allows easy swapping of executables, simplifying maintenance and resource optimization.
• Data-level parallelism in Swift/T handles multiple samples efficiently, though debugging can be more complex.
• Novel language features were integrated into the open-source repository.

Conclusions:

• Swift/T provides a powerful and portable solution for scaling bioinformatics analyses.
• The system facilitates efficient management of complex, large-scale genomic data workflows.
• The developed criteria and implementation offer a blueprint for robust, production-ready bioinformatics workflows.