Fig. 2

Transitional states of cancer cells revealed by trajectory mapping. A UMAP embedding of cancer cells overlaid with unsupervised cluster cell type annotations (left), proportional sample contributions to each cell type cluster (middle), and sample label (right). B Pseudotime trajectory of ESCC cancer cells in a two-dimensional state space inferred by the “Monocle 2” method. C Cancer cells mapped to the branched structure in the trajectories (left) and the distribution of cancer clusters in ESCC tumors stratified by treatment and pathological response (right). D Cell number count in cancer clusters identified in ESCC tumors of patients before (T_B) and after (T_A) neoadjuvant chemo-immunotherapy. E Pseudotime trajectory of ESCC cancer cells showing three differentiation states. F Investigation of biological processes (BP) through Gene Ontology (GO) pathway enrichment analysis in cancer cells across three differentiation states. G Heatmap shows the top 500 genes with significant autocorrelation grouped into 12 gene modules based on pairwise correlations of gene expression in cancer cells. H The scatter plots show the expression of the top five highly expressed regulons in each of the six cell subsets. I Heatmap of the area under the curve (AUC) scores of expression regulation by transcription factors in ESCC tumors stratified by treatment and pathological response, estimated by SCENIC