Fig. 2

Recovery of structural relationships in the in silico tissues. A Voronoi diagram representation of the generated in silico tissue structures: Tissue 1—random structure, Tissue 2—structure with self-preference of a single cell type and Tissue 3—structure with mutual preferences of two cell types. B Amount of variance explained (percentage points) of the identity of the cell type when taking into account the information about the distribution of cell types in the immediate neighborhood. C Estimated importance of the relationships from the cell type distribution in the immediate neighborhood. Tissue 1 is missing as there was no information captured about the structure of the random tissue. Some of the target cell types are missing as the heatmaps contain only those targets with gain of variance explained of more than 5% and importances larger than 1 (one standard deviation above the mean importance of all predictors for that target). D Estimated importances of the expression of genes in Tissue 3 coming from the immediate neighborhood as predictors of the expression of the target gene. Shown are target genes with variance explained above 4%, highlighted are the interaction with estimated importance of 2 and higher. The genes shown in bold are either top 10 markers of ct0 or ct2. There is only one gene (gene_33) that is in the top 10 markers of the other two cell types. E Distributions of the estimated importance of the predictor-target relationship from the juxtaview of the top 10 markers of the individual cell types to the markers of cell types 1 (above) and 3 (below) in Tissue 3