Containers for computational reproducibility

The fast-paced development of computational tools has enabled tremendous scientific progress in recent years. However, this rapid surge of technological capability also comes at a cost, as it leads to an increase in the complexity of software environments and potential compatibility issues across systems. Advanced workflows in processing or analysis often require specific software versions and operating systems to run smoothly, and discrepancies across machines and researchers can impede reproducibility and efficient collaboration. As a result, scientific teams are increasingly relying on containers to implement robust, dependable research ecosystems. Originally popularized in software engineering, containers have become common in scientific projects, particularly in large collaborative efforts. In this Primer, we describe what containers are, how they work and the rationale for their use in scientific projects. We review state-of-the-art implementations in diverse contexts and fields, with examples in various scientific fields. Finally, we discuss the possibilities enabled by the widespread adoption of containerization, especially in the context of open and reproducible research, and propose recommendations to facilitate seamless implementation across platforms and domains, including within high-performance computing clusters such as those typically available at universities and research institutes. Containers are increasingly used in scientific projects to ensure reproducibility and promote collaboration. In this Primer, Moreau et al. explain containers, their use for scientific research and across various fields.