Extended Data Fig. 2: Cell state transition dynamics predict phylogenetic correlations.

a, Simulated idealized phylogeny containing 2⁶ = 64 cells (Supplementary Note) in which cells can transition between three possible cell states. Cell state transitions are represented as a discrete-time Markov chain (Supplementary Note). b, Simulated cell state transition dynamics and measured phylogenetic autocorrelations (Methods) for the first cell state for 1,000 independent simulations on idealized phylogenies, containing 2¹⁰ cells, in which state transition probabilities were randomly generated for each trial. Phylogenetic correlations were computed using a weighting function that included only sibling cells (one-node only, as described in Methods). LOESS regression line (blue) with 95% confidence interval (light gray) is shown. c, (Left) Simulated versus PATH-inferred cell state self-transition (that is, stability) probabilities (Methods), by transforming the phylogenetic autocorrelations measured in b. (Right) Simulated versus PATH-inferred cell state transition probabilities from state 1 to 2, on idealized phylogenies (Supplementary Note). Dashed red lines both have slope 1 and pass through the origin. Linear regression lines (blue) with 95% confidence intervals (light gray) are shown for both plots. Spearman’s correlation coefficients shown in panels b and c had two-sided P values < 2.2e-16, the limit of precision in R.