Fig. 2From: Griottes: a generalist tool for network generation from segmented tissue imagesa Confocal dorsal view of the zebrafish adult pallium. Glial fibrillary acidic protein (GFAP) in green, Proliferating cell nuclear antigen (PCNA) in magenta and Zonula occludens-1 (ZO1) in white or blue. ZO1 is highlighting the apical domain of the cells allowing the identification of their apical area. Inset: spotlight on a limited tissue area, the ZO1 membrane staining allows for the exact localization of cell membranes. Scale bar is 100 μm. b Griottes incorporates different network construction methods. The contact-based method connects cells sharing a common membrane, the Delaunay and Geometric graphs are commonly used graph-generation methods. c The graphs generated from a same set of nodes with different construction rules have different properties. For instance, the degree distribution of a Geometric graph is broader than that of the Contact and Delaunay graph. d Mean PCNA signal within the cells in the example tissue. Cells with an average intensity above 6500 (red line) are considered PCNA+, the other cells are PCNA−. e Thresholding intensity signals converts a network populated with continuous fluorescence signals to a network populated with categorical cell types. f. Representation of the example network (panel b) where node colors represent cell type. Left: the network is projected on the ZO1 signal. Right: the network is projected on the PCNA signal. This method reliably incorporates cell type information into the network representation of the tissue. g Left: connected cells can have widely varying contact surfaces. Right: this information can be encoded into the network by weighting the links between cells. Two differing cell-cell interfaces (pink lines) have different link weights in the network representation of the tissue (pink arrows). h The connection between cells can be quantified at scale: the histogram of link weights in the tissue shown in panel aBack to article page