Answers are grounded in this paper and cite their sources.
Introduction: Clinical and translational research increasingly depends on multidisciplinary teams with specialized data expertise, yet the term data science carries markedly different meanings across fields and institutions. This conceptual ambiguity affects team science, training, and research rigor. We propose a shared framework for the biomedical data sciences: biostatistics, epidemiology, bioinformatics, informatics, and computational biology, grounded in foundational principles common across disciplines. Methods: We trace the historical evolution of the biomedical data sciences, characterize heterogeneity in training across disciplines, and present survey evidence from ACTS BERD and Informatics Special Interest Group members ( n = 88, August 2025) documenting variation in how the term is used and who claims it. Results: Survey respondents varied widely in whether they identified as data scientists and how they defined data science, even within a community of practitioners sharing institutional homes. We interpret this variation not as confusion to be corrected, but as evidence that the singular term obscures distinct disciplines with a broad spectrum of academic traditions. This motivates our proposal: a framework built on what these disciplines share rather than what divides them. Conclusions: The heterogeneity of the biomedical data sciences is a feature of a maturing field, not a problem to be eliminated. A shared framework built on common foundational principles, including matching scientific questions to: appropriate analytic goals, study design, uncertainty quantification, bias mitigation, and reproducibility, offers a path toward more intentional training, earlier engagement of data expertise, and stronger interdisciplinary collaboration in clinical and translational research.
No takes yet. Share an insight, caveat, or question.
Messinger et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: