From: Where next for the reproducibility agenda in computational biology?
Software developers can: |
• Use well thought-out and appropriate principles of modularity in designing software. • Provide practical, comprehensive advice on installation. Check it by installing software on commonly-used systems, or simplify it using a platform such as Docker. • Provide code annotation and manuals in multiple, accessible forms with different levels of detail. |
Computational researchers can: |
• Make use of dynamic and/or updatable formats for publishing research, where appropriate. • Ensure they provide all details of how an analysis was carried out, including providing all the code and data necessary to reproduce a result. Context-specific minimum information guidelines can provide useful checklists. • State explicitly what are the key features of a piece of published work, how to measure agreement when the work is reproduced, and how close the agreement is expected to be. |
The computational biology community can: |
• Introduce a “seal of approval” for good reproducibility practice including adherence to reporting checklists, which could be awarded to labs, individual researchers or particular pieces of software or research. • Require adherence to appropriate minimum information checklists for publication in peer-reviewed journals and through other channels. • Promote and campaign for education in good computational practice for scientists of all backgrounds, from undergraduate to professorial level. • Provide structures and opportunities for networking, support and professional development of computational researchers. |