Computational Reproducibility: A Practical Framework for Data Curators

Abstract

Introduction: This paper presents concrete and actionable steps to guide researchers, data curators, and data managers in improving their understanding and practice of computational reproducibility. Objectives: Focusing on incremental progress rather than prescriptive rules, researchers and curators can build their knowledge and skills as the need arises. This paper presents a framework of incremental curation for reproducibility to support open science objectives. Methods: A computational reproducibility framework developed for the Canadian Data Curation Forum serves as the model for this approach. This framework combines learning about reproducibility with recommended steps to improving reproducibility. Conclusion: Computational reproducibility leads to more transparent and accurate research. The authors warn that fear of a crisis and focus on perfection should not prevent curation that may be ‘good enough.’

Reproducibility in materials informatics: lessons from ‘A general-purpose machine learning framework for predicting properties of inorganic materials’

“…A signicant hurdle to reproducibility emerges when developers presume that dependent soware packages will remain compatible and readily accessible for years to come. However, the challenges posed by evolving, superseding, or abandoned dependencies are a well-recognized issues in computational reproducibility 27,28 with many potential solutions, each accompanied by drawbacks. Docker, for instance, has gained popularity as an open platform for sharing and running tools within a loosely isolated environment to circumvent issues related to dependency management.…”