fair-pipelines-poster (1).pdf (796.51 kB)

A FAIR Approach to Neuroimaging Analysis with Boutiques

Download (796.51 kB)
posted on 19.05.2019, 13:06 by Greg Kiar

In recent years the FAIR principles have become a guide for sharing scientific data. Products that implement these principles should be Findable, Accessible, Interoperable, and Reusable. While these guidelines have often been interpreted specifically with data in mind, they can be meaningfully applied to other scientific products such as tools or software. While many database systems and organizational standards are evolving to implement these principles for datasets, the tool landscape remains cloudy.

Platforms such as GitHub allow users to access software, but tools can often be difficult to find and platforms remain mainly suitable for open-source projects. Standards for automatically generated documentation increase the re-usability of software but are unlikely to survive beyond a tool’s supported lifetime. Common frameworks for handling tool arguments provide a consistent interface to tools, though these frameworks vary across libraries. While virtualization engines such as Singularity encapsulate tools and their environments for their re-use, the barrier often remains high to evaluate the performance of a tool within one’s workflow.

We present an end-to-end approach for producing and consuming FAIR tools:

  • Findability is achieved through the public indexing and minting of Digital Object Identifiers (DOIs) for executable tool descriptions.

  • Accessibility is facilitated through the accompaniment of tools and both human-readable and executable documentation.

  • Interoperability is gained through the adoption of a common and versatile standard for these tool descriptions, Boutiques, and a common application programming interface (API) for medical imaging, CARMIN, which provides a standard set of instructions for accessing data processing services.

  • Re-usability is facilitated through integrated testing, simulation, and adoption of these descriptions in high-performance computing (HPC) environments.