A how-to guide for code-sharing in biology (2401.03068v1)

Abstract: Computational biology continues to spread into new fields, becoming more accessible to researchers trained in the wet lab who are eager to take advantage of growing datasets, falling costs, and novel assays that present new opportunities for discovery even outside of the much-discussed developments in artificial intelligence. However, guidance for implementing these techniques is much easier to find than guidance for reporting their use, leaving biologists to guess which details and files are relevant. Here, we provide a set of recommendations for sharing code, with an eye toward guiding those who are comparatively new to applying open science principles to their computational work. Additionally, we review existing literature on the topic, summarize the most common tips, and evaluate the code-sharing policies of the most influential journals in biology, which occasionally encourage code-sharing but seldom require it. Taken together, we provide a user manual for biologists who seek to follow code-sharing best practices but are unsure where to start.

• The paper presents actionable guidelines to enhance code sharing among biologists, improving reproducibility and transparency.
• It examines journal policies and finds that only 10% require analytical code sharing, despite widespread data deposition mandates.
• It advises sharing complete scripts, configuration files, and dependency lists to ensure replicable and verifiable research.

Recommendations for Effective Code Sharing in Computational Biology

The paper "A how-to guide for code-sharing in biology" addresses the pressing need for effective strategies to share computational code within the biological research community. As computational biology continues to intersect with traditionally wet lab disciplines, the integration of computational methodologies becomes increasingly vital. This paper provides comprehensive guidelines aimed at enhancing transparency and reproducibility in scientific research by focusing on best practices in code sharing.

Key Insights and Contributions

The core of the paper is structured around a set of recommendations tailored for biologists who may not be intrinsically familiar with computational practices but need to engage in open science initiatives. The authors diligently review the current landscape of code-sharing policies in prominent biological journals, highlighting a significant gap: while a majority mandate data sharing, a mere 10% require the disclosure of analytical code.

Insights from Journal Policy Survey:

• Out of 100 prominent journals, only 10 enforce mandatory sharing of analytical code, even as 55% require the public deposition of sequencing data.
• Journals such as Science, Nucleic Acids Research, and Radiology are cited as leading the requirement for public accessibility of code, although similar expectations do not uniformly extend across all publications.

Code Sharing Recommendations:

The paper suggests specific types of code files that should be shared, including:

• Scripts for data-cleaning, analysis, and data visualization to enable reproducibility and transparency.
• Parameters and configuration files used in command-line operations, ensuring processes can be reliably replicated.
• A complete list of dependencies to account for variations in software environments and versions that might affect computational outcomes.

The authors offer pragmatic solutions for code preparation, suggesting that sharing imperfect but functional code is substantially better than withholding it entirely. The emphasis is on striking a balance, wherein code serves to demonstrate a computational approach specific to a single paper rather than expecting universal applicability.

Implications and Future Directions

From a practical perspective, the paper underscores that transparency in code sharing mirrors the rigor expected in traditional experimental methodologies. By adopting these recommended practices, researchers can contribute to an ecosystem where computationally derived conclusions are more readily verifiable and extendable by others.

On a broader theoretical level, the paper aligns with ongoing discussions about open science. There is a need to transition the cultural acceptance of code production and sharing from optional to a norm in the lifecycle of scientific work. This could potentially lead to standardized practices across different biology subfields, thus alleviating some of the discrepancies observed in policies at interdisciplinary intersections.

Speculation on Future Developments

Looking forward, the biological research community could benefit significantly from technology-driven solutions that automate some aspects of code sharing and documentation. For instance, integration with advanced workflow platforms could facilitate seamless sharing and version control of analysis pipelines. Furthermore, policy shifts within academic journals towards stronger mandates on code sharing could drive widespread adoption of these practices, fostering an environment of enhanced scientific integrity.

In conclusion, this paper provides vital direction on addressing the current limitations in code sharing among biologists. By advocating for comprehensive sharing practices, the authors contribute to the foundational shift towards increased reproducibility and openness in scientific research.