A method to track dataset reuse in biomedicine: filtered GEO accession numbers in PubMed Central

摘要

Reusing research data has important potential benefits:rngenerative science and efficient resource use. Tracking thernreuse of research datasets would allow us to understandrnwhether the potential benefits are indeed realized, enablernrecognition of investigators who produce, annotate, andrnshare useful data, and inform data sharing and reuserninitiatives, tools, and policies.rnUnfortunately, the lack of clear attribution practices for datarnmake automated tracking of data reuse difficult. I present arnmethod for tracking research data reuse that takesrnadvantage of the community norms around gene expressionrnmicroarray data sharing and the rich NCBI Entrezrnresources. Specifically, the full-text of papers stored inrnPubMed Central are queried for accession numbers ofrndatasets archived in NCBI’s Gene Expression Omnibusrn(GEO) repository. Studies known to have createdrnmicroarray data are excluded through automated filters andrnguided manual curation. MeSH terms attached to the datarncreation and data reuse studies provide additionalrninformation for analysis. Finally, I extrapolate the findingsrnto all of PubMed.rnAutomated portions of this method have been implementedrnin python and are openly available. Although imperfect,rnthis dataset is a valuable initial resource for research intornpatterns of data reuse.