Weill Cornell Medicine Samuel J. Wood Library's Response To The Biomedical Research Reproducibility Crisis

Background: The biomedical research field is more and more frequently reporting a reproducibility crisis. A 2016 Nature survey showed 70% of researchers in biology were unable to reproduce other scientists’ findings and that 60% were unable to reproduce their own findings (Baker, https://doi.org/10.1038/533452a). Common factors contributing to the lack of reproducibility that can readily be acted upon include lack of access to methods details and raw data, poor reporting methods, poor data management, and inability to manage complex datasets. The Weill Cornell Medicine Samuel J. Wood library is mounting a new service named “Research Reproducibility Support Service” (RRSS) to foster best practices in biomedical research reproducibility and improve compliance with funder policies.

Description: The RRSS started in 2021 when the organization’s Research Integrity Office asked the library to develop a methodology for identifying image manipulation. It now includes consultations for bioinformatic pipeline automation and large datasets analysis; development of an Image Forensics web tool allowing patrons to check for duplicated or falsified scientific images; and development of instructional materials on best practices in biomedical research reproducibility (direct computational and analytic replication). A closed, operational, beta version of the Image Forensics web tool already exists. However, this version needs further improvement in terms of detection accuracy. In addition, a diversification of the type of scientific materials that can be analyzed will also be implemented. Consultations for bioinformatic pipeline automation and teaching material for research reproducibility best practices are current and operational.

Conclusion: Our program evaluation will encompass the accuracy of detecting manipulated images, effectiveness of direct computational and analytic replication techniques on big data sets, and patron confidence in the robustness of their methodology.