Flexible & Executable Provenance in Data-Intensive Biomedical Research: A Flexible Research Data Service

Duke University, Durham NC

Investigators

Linked publications & trials

Abstract

? DESCRIPTION (provided by applicant): Tracing the lineage of scientific data and assertions is critical for the checks and balances that once ensured scientific fidelity (Collins & Tabak, 2014). As data become pervasively digitized, generating and following lineages automatically and at scale increases the usefulness and quality of conclusions. A significant challenge in data-intensive science is generating that lineage-the provenance-of scientific information, while facilitating retrieval and re-execution. We hypothesize such capabilities improve the reproducibility of assertions and make data more useful to society. Our objective is to build application programming interfaces for provenance, data-integrity, storage, and reproducible workflows that empower researchers to record, retrieve, and re-run scientific lineages. The rationale for the proposed research is that the value of the scientific data is enhanced by being able to retrospectively reproduce a result and by understanding its origins for future use. Provenance also facilitates measurement of data's importance-its impact. Guided by strong preliminary work, we will test our hypothesis by pursuing two specific aims: (1) Building APIs for provenance, data management, data integrity, and re-executable workflows, (2) Providing a platform for storing and deploying containerized compute environments that also serves as a learning laboratory for reproducible data science. This approach is innovative in focusing on flexibility and accommodating the myriad use cases across biomedical science, while pro- viding a hub for training investigators in reproducible data science. By creating an open source Flexible Re- search Data Service, the proposed research will significantly impact our ability to make our investments in biomedical research more useful.