Dockstore: The Community Platform for Reproducible Biomedical Workflows and Applications

点击翻译按钮获取中文摘要

Dockstore - A Platform for Creating, Sharing, Publishing and Reproducing Computational Science For science to work analyses, tools and workflows must be reproducible; reproduction is essential to building consensus and confidence, and improving the state of the art. A large and growing set of scientific methods are computational in nature, and we can make these computational methods exactly reproducible. Dockstore is a repository for scientific tools and workflows that provides the means precisely to capture the code, parameters and operating system environment that are necessary. For tool users it makes it easy to find a given tool and then run it precisely as the author intended. It is possible to export a tool for use in a cloud platform, such as NHGRI AnVIL, with a few clicks, or similarly easily download it to an institutional compute environment, or even a laptop. For tool creators it provides an integrated system to publish versioned scientific tools with a citable Digital Object Identifier (DOI), linking to containers, existing code repositories and capturing vital parameter configuration information. Dockstore, previously funded by an R01 from NHGRI grant, has grown tremendously over the last four years. It now has thousands of users, hosts hundreds of tools and workflows, and is used as the methods for multiple major platforms and projects. Though not technically a renewal, this proposal would provide the core funding to continue the development and growth of Dockstore, which is a fully open source platform entirely reliant on grant funding. This proposal would grow Dockstore as a resource to make it a hub for broadly sharing computational science. The aims of the proposal cover foundational infrastructure development (Aim 1), support for sharing new, popular content types (Aim 2), the expansion of publishing features to make Dockstore more broadly useful (Aim 3), and a strong focus on training and outreach (Aim 4). The infrastructure development of Aim 1 will increase the scaling and robustness of Dockstore and support Dockstore as a secure, compliant resource, enabling it to safely host pre-publication tools and workflows working with managed access datasets. It will integrate Dockstore more deeply with secure cloud platforms, such as the NHGRI AnVIL. The expansion of content types by Aim 2 will, for example, make it possible to publish programming notebooks on Dockstore. Such artifacts are widely used but not yet robustly searchable by any scientific resource. Aim 3 will bring ORCID integration, allowing users to authenticate themselves and then sign and endorse their work, and to provide open peer review of published tools. Aim 4 will focus on growing and educating the user base via an expansion of online training material, webinars, in person workshops and collaborations with training for partner projects using Dockstore. We will also further develop means to continuously acquire user feedback.