Answers are grounded in this paper and cite their sources.
Single-cell technologies, coupled with powerful machine learning, have multiplied the number of named innate immune cell states faster than their biological features have been established. Innate cell identity is generated by at least four partially independent axes (ontogeny, maturation, response to microbial or damage signals, and tissue adaptation) that project onto a shared transcriptional space. Consequently, snapshot expression data alone cannot uniquely attribute an observed partition to any one of these processes. Data-driven taxonomy has been particularly effective where one axis provides a stable interpretative anchor, as in ontogeny-defined conventional dendritic cells, and faces greater challenges where several axes interact, as in granulocytes, tissue-adapted macrophages, and ILCs. Here, we assess what recent atlas and foundation model studies have resolved, identify independent biological validation rather than computational resolution as the current bottleneck, and propose an explicit evidence hierarchy for deciding when computational clusters should be considered bona fide biological entities.
No takes yet. Share an insight, caveat, or question.
Bonavita et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: