Fig. 4

CCC pathways distinguishing immune responses to pathogenic and nonpathogenic lentiviral infection. A Overview of application of Scriabin [42] to analyze single-cell resolution CCC profiles. Briefly, Scriabin’s summarized interaction graph workflow was used to align the time points within an animal and identify cell type combinations with perturbed CCC to prioritize for downstream analysis. Then, Scriabin’s ligand activity ranking process was used to identify biologically active CCC edges. Finally, Scriabin’s CCIM workflow was used to visualize the full structure of CCC phenotypes (see “Methods”). B Scatter plots depicting the degree of CCC perturbation predicted by Scriabin’s summarized interaction graph workflow (y-axis). The most highly perturbed cell type combinations for each animal are labeled (sender cell type — receiver cell type). Points are sized by the variability in perturbation score between bins of the same cell type combination. C Plot depicting the distribution of time points at which CCC in each monocyte-to-monocyte bin-bin pair is maximally perturbed. D UMAP projections of the most highly perturbed monocyte-monocyte cell-cell pairs from all samples, colored by animal (left) and time point (right). E Stacked bar plot depicting DE ligand-receptor pairs in the monocytes of each animal at the time point of maximal perturbation. A ligand-receptor pair is considered to be more differentially expressed in an animal class if > 1 standard deviation of that ligand-receptor pair’s log(fold-change) is contributed by a particular animal class. F Predicted ligand activities of monocytes over time in each animal. G Alluvial plots depicting ligand-target gene connections within monocytes of each animal at the time point of maximal perturbation. H UMAP projections of the most highly perturbed CD14 monocyte-CD8 T_EM sender-receiver cell-cell pairs colored by animal (left) and clusters enriched in 170-infected animals (right). I Bar plots depicting DE ligand-receptor pairs in the clusters highlighted in H