Fig. 5: Computational lineage modeling allows to infer a robust trajectory of tumorigenesis.
From: Tracing tumorigenesis in a solid tumor model at single-cell resolution
a Diffusion map of tumor-specific epithelial cell populations together with inferred trajectory obtained by smoothing diffusion coordinates over pseudotime. b Density plot of cells from the different subpopulations along pseudotime. c Normalized EMT, Wnt and stemness64 pathway scores, smoothed over pseudotime using LOESS regression (gray shading indicates 95% confidence intervals). d Density plot of cells from double-mutant P40 and P90 samples with ≥5 cells in the 4 subpopulations along pseudotime. P-value from linear model using 855 cells in 12 samples. e Proportions of the subpopulations from the same samples according to stage (P-values from binomial mixed effects model). Boxes span the 25th to 75th percentile, whiskers 1.5 times the interquartile range. f Heatmap of SAVER-imputed gene expression for the top 100 differential genes between the 4 subpopulations (top 16 are highlighted).
