Fig. 4: Niche-phenotype mapping reveals two fundamental properties of tissue phenotypic architecture.
From: NIPMAP: niche-phenotype mapping of multiplex histology data by community ecology

a–d The spatial context of cells is a stronger determinant of phenotype than cell-autonomous effects. Dendritic cells are illustrated here and other cell types appear in Supplementary Fig. 8d, e. b Niche-phenotype mapping (left) and (spatially-agnostic) phenotypic clusters (right) identify common phenotypic markers—CD45RO, Keratin 6—but also divergent markers—HLA class 1, HLA-DR, CD138. b If cell-autonomous effects determine phenotypes more than spatial context, spatial context will poorly associate with phenotypes. Thus, phenotypic clusters—which are built independently of spatial context—will predict a cell’s niche as poorly as a random predictor. The sensitivity and specificity of phenotypic clusters are computed using cluster membership as a predictor of niche membership. c If the spatial niche context of a cell is the main determinant of phenotypes, phenotypes will strongly associate with niches. Thus, the niche of a cell can be predicted from a cell’s phenotypic cluster. d The data supports the hypothesis that the spatial context of cells is a stronger determinant of phenotype than cell-autonomous effects. e Both niches and their interfaces structure the spatial architecture of phenotypes. The contribution of a niche or interface to phenotypic architecture is quantified by summing the squared correlations of all phenotypes with that niche (rows of the matrix shown in Fig. 3a). Squared correlations are expected to be higher for niches and lower for interfaces in a scenario where niches structure phenotypic architecture (left). Conversely, under the hypothesis that phenotypic architecture is structured by interfaces, squared correlations will be higher for interfaces and lower for niches (middle). The data shows that both niches and interfaces have high squared correlations and thus contribute to phenotypic architecture (right). TLS: tertiary lymphoid structure. Source data are provided as a Source Data file.