Fig. 4: Distinct evolutionary trajectories identified from single-cell transcriptomics explain functional heterogeneity in chronic infections.
a Schematic of the method to identify differentiation trajectories from scRNA-seq data (created with biorender.com) utilising partition-based graph abstraction (PAGA), a priori information on longitudinal samples and IFN-γ ELISpot data. b PAGA graphs of single cells (n = 903) identifying 13 clusters and their probability of connection (thickness of lines). c Distribution of the estimated days post-infection (DPI) corresponding to the sample time point associated with each cell. d PAGA graphs (visualised using ForceAtlas2 layout algorithm (FA1, FA2)) coloured by epitope specificity, disease outcome, disease stage, magnitude of IFN-γ response, rate of escape (slow, fast, no escape), and clone size (log-transformed). e Distribution of T cell phenotypes across PAGA clusters (as per b). Single-cell phenotypes are obtained via index sorting protein expression data. f Trajectories derived from scRNA-seq data with coloured pseudotime values. Loess curves represent the changes along the trajectories of both single-cell protein and gene expression values. The 95% confidence intervals are denoted by shaded regions. g Dot plot of selected differentially expressed genes (derived from pairwise comparisons with a two-sided hurdle-model from MAST) between the early and late phases of each trajectory (p < 0.05, |log2(FC)| ≥ 1, coloured). Dot size represents the proportion of cells with non-zero expression. Early ≤120 DPI, late >120 DPI. h Enriched pathways identified from GSEA using differentially expressed genes (as per (g)) between early and late phases within each trajectory. Adjusted p values <0.05 in at least one trajectory phase. NES: normalised enrichment score.
