Fig. 3: MuTrans reveals the cell-state transitions in EMT dataset of Squamous Cell Carcinoma.

a Attractor basins found by MuTrans, with cells plotted in the UMAP low-dimensional space and colored by MuTrans attractors. b Calculated transition entropy of each cell, with larger values corresponding to more intermediate cells with highly mixing identities. c, d Gene expression value of epithelial state (E) marker Epcam and mesenchymal state (M) marker Mmp19. e The dynamical manifold of EMT with cells colored by MuTrans attractors in (a). Two dashed squares indicate the transition processes presented in (h) and (i). Attractor 0 is annotated as the epithelial (E) state, attractor 0 and 4 as two mesenchymal (M) states, and attractor 1 and 2 as two intermediate cell states (ICS) of EMT. f, g The transition path analysis by setting E attractor as start state and two M attractors as target state, respectively, overlaid on the dynamical manifold. The numbers indicate the relative likelihood of each transition path. The cells are colored by MuTrans attractors. The grayness indicates energy values of dynamical manifold, with darker colors representing lower energy values. h, i Transition cells and genes analysis during the transitions from E to ICS2 (h) and from ICS2 to M2 (i). For each cell-state transition, the cells (columns) of heatmap are ordered by transition cell scores (TCS) of each process, and transition cells marked by the black dashed rectangles. The rows of heatmap are the top five most significant genes identified for the down-regulated MS (meta-stable) genes, down-regulated IH (intermediate-hybrid) genes, TD (transition-driver) genes, up-regulated IH genes and up-regulated MS genes, respectively (from top to bottom in heatmap).