Fig. 9

Results of the CellPhoneDB analysis on our scRNA-seq dataset. A 81,893 significant interactions were retained based on predicted p values for a ligand–receptor complex across two cell clusters, calculated using permutations in which cells are randomly re-assigned to a cluster. The strongest interactions, based on the number of interactions found, were shown between fibroblasts, tumour cells and endothelial cells. B Graph network representation of the interactions between subclusters. Only subcluster pairs with more than 170 interactions (i.e. 70% of pairs) are displayed. Edges are weighted by number of interactions. On a subcluster level the strongest interactions are found between FB_COL27A1, FB_SERPINE1 and FB_COMP. C Table illustrating the fraction of the interactions involving the different cell major cell types across the molecular subtypes. Dendritic cells showed scarce interactions in mesenchymal HGSTOC (3% compared to 16–24% in the other molecular subtypes) but as only 1 DC was detected in patient 6 the results for DCs in mesenchymal HGSTOC are not reliable. D Heatmaps visualising the distribution of the significant interactions between the different cell phenotypes across the different molecular subtypes: patient P1 proliferative HGSTOC, patient P2 differentiated HGSTOC, patient P5 immunoreactive HGSTOC and patient P6 mesenchymal HGSTOC. Zoom-in on the molecular subtype-specific interactions. Intensity of the interactions is measured is scaled by with the number of interactions between those two subclusters