Making Biomedical Research Software FAIR: Actionable Step-by-step Guidelines with a User-support Tool

Research software is increasingly crucial in scientific research, with surveys indicating its widespread use and indispensability. Its importance is particularly evident in biomedical research, fueled by advancements in machine learning and artificial intelligence. The FAIR (Findable, Accessible, Interoperable, and Reusable) principles, while valuable for data, require adaptation for research software. The FAIR4RS Working Group addressed this need, publishing revised FAIR principles tailored for software. However, these principles remain aspirational, lacking practical instructions for researchers. Existing guidelines offer some actionable steps but are either incomplete or based on older, less comprehensive FAIR principles. This research fills this gap by providing the first minimal, actionable, step-by-step FAIR-BioRS guidelines specifically for biomedical research software, developed through an in-depth study of FAIR4RS principles and a review of current practices. To ease implementation, a user-friendly workflow is integrated into the FAIRshare application.

The authors conducted a comprehensive literature review to inform the development of the FAIR-BioRS guidelines. They examined publications on FAIR principles for research software, resources cited within those publications, relevant materials from the FAIR4RS Working Group, and PubMed searches focusing on FAIR and research software reusability. The review encompassed general best practices in software development and specifically those relevant to biomedical research software. Key areas explored included coding standards, metadata practices (e.g., CodeMeta, CITATION.cff), licensing, appropriate repositories (archival and deployment), and relevant registries (bio.tools, RRIDs). The review highlighted a lack of community consensus on various aspects, such as standard metadata formats and identifiers for software, underscoring the need for the FAIR-BioRS guidelines.

The methodology involved a multi-step process. First, concise, high-level instructions were derived for fulfilling each of the 17 FAIR4RS principles. These instructions were then categorized into five action-based themes: developing software following standards and best practices; including metadata; providing a license; sharing software in a repository; and registering in a registry. Next, for each category, outstanding questions were identified to guide the literature review. The review, detailed in a PRISMA diagram, involved analyzing 39 relevant resources. The findings were synthesized, incorporating the authors' expertise and external feedback, particularly where community consensus was lacking, to generate recommendations. These recommendations were then organized into a step-by-step FAIR-BioRS guideline workflow that aligns with typical software development processes. The FAIRshare application was developed to support implementation of these guidelines, integrating user interfaces, automation, and helpful resources.

The FAIR-BioRS guidelines, presented as a six-step process (prepare prior to development; follow coding standards and best practices; document software; include metadata files; share software in a repository; register software in a registry), provide actionable steps for making biomedical research software FAIR. Recommendations included using version control systems (GitHub, GitLab, Bitbucket), employing appropriate coding standards (PEP 8 for Python, Google's R Style Guide), documenting code and dependencies thoroughly (README, CHANGELOG), including CodeMeta.json and CITATION.cff metadata files, choosing an OSI-approved open-source license (MIT or Apache 2.0 recommended), archiving software on Zenodo or Figshare, and registering in bio.tools and optionally RRIDs. The FAIRshare application simplifies this process by guiding users through the guidelines with an intuitive interface, automating tasks like metadata generation and facilitating sharing on Zenodo and Figshare. FAIRshare addresses challenges like obtaining DOIs before publication to ensure proper inclusion in metadata. The guidelines align with NIH recommendations and have incorporated community feedback through various presentations and discussions.

The FAIR-BioRS guidelines address the significant gap in providing actionable steps for making biomedical research software FAIR. While aligning with NIH recommendations, they provide a more detailed and complete approach to achieving FAIR compliance, specifically addressing the nuances of research software. The development process incorporated extensive community feedback, and the resulting guidelines and the supporting FAIRshare application aim to increase the adoption of FAIR practices within the biomedical research community. The authors identify ongoing needs for community consensus, including standardized metadata practices, interoperable data formats, and a common identifier for software. The FAIR-BioRS guidelines and FAIRshare application are intended to evolve with community input and evolving standards.

The FAIR-BioRS guidelines offer a much-needed practical framework for biomedical researchers to develop and share FAIR research software. The accompanying FAIRshare tool further simplifies this process. Future work includes creating more specific guidelines for different software types (e.g., Python packages, Jupyter notebooks) and expanding FAIRshare's capabilities to encompass other repositories and automation features. The open-source nature of both the guidelines and FAIRshare promotes community engagement and continuous improvement.

The FAIR-BioRS guidelines currently focus on biomedical research software. While many aspects may be adaptable to other fields, specific recommendations may need adjustments. FAIRshare, in its current version, assists in preparing for publication but doesn't integrate directly into the development process. Future development will address this limitation by integrating directly with version control systems.