Analytical code sharing practices in biomedical research

Area of Science:

• Biomedical research
• Computational biology
• Data science

Background:

• Data-driven computational analysis is vital in modern biomedical research.
• Lack of shared research outputs (data, code, methods) impedes study transparency and reproducibility.
• Many published studies are irreproducible due to insufficient sharing of documentation, code, and data.

Purpose of the Study:

• To analyze the extent of code sharing in biomedical research publications.
• To identify factors associated with code availability and reproducibility.
• To propose strategies for improving code sharing practices.

Main Methods:

• Comprehensive analysis of 453 biomedical manuscripts published between 2016-2021.
• Assessment of code availability, data sharing, and code organization.
• Statistical analysis to identify associations between code availability statements, analysis type, and code sharing.

Main Results:

• 50.1% of analyzed manuscripts failed to share analytical code.
• Most studies sharing code did not share additional research outputs like data.
• Only 10% of articles organized code in a structured, reproducible manner.
• Code availability statements were significantly associated with increased code availability.
• Studies performing secondary analyses were more likely to share code than primary analyses.

Conclusions:

• There is a significant deficit in code sharing practices within biomedical research, impacting reproducibility.
• Promoting code availability statements and structured code organization is essential.
• Prioritizing open science practices, including code and data sharing, is necessary to enhance scientific rigor, collaboration, and accelerate discoveries.