This paper proposes an original definition of samples as a kind of data within the relational framework of data. The distinction between scientific objects (e.g., samples, data, models) often needs to be clarified in the philosophy of science to understand their role in the scientific inquiry. The relational framework places data at the forefront of knowledge construction. Their epistemic status depends on their evaluation as potential evidence in a research situation and their ability to circulate among researchers. While samples are significant in data-generating science, their role has been underexplored in the philosophy of data literature. I draw on a case study from data-centric microbiology, viz. amplicon sequencing, to introduce specifications of the relational framework. These specifications capture the distinctive epistemic role of samples, allowing the discussion of their significance in the inquiry process. I argue that samples are necessarily transformed to be considered as evidence, portable in the limits of a situation, and they act as world anchors for claims about a phenomenon. I compare these specifications with other data and evidence frameworks and suggest they are compatible. The paper concludes by considering the extension of these criteria in the context of biobanking. The specifications proposed here help analyze other life sciences cases and deepen our understanding of samples and their epistemological role in scientific research.