Extended Data Fig. 4: PATHpro inferences from simulations of somatic evolution with cell state-specific proliferation rates.

a, Comparison of PATH versus PATHpro transition inference errors (Euclidean distance between inferred and simulated transition probabilities) on simulated data when proliferation rates depend on cell state (Methods). Phylogenies were simulated in forward time, starting with one cell and ending after reaching 10⁴ cells, and then subsampled to 10³ cells. The state of the first cell was randomly chosen, and \({\gamma }_{1}\), the proliferation rate of state 1, is shown by the x-axis, and \({\gamma }_{2}\), the proliferation rate of state 2, was fixed at 1. Cell state transitions between states were symmetric with \({P}_{12}={P}_{21}=0.1\). 100 phylogenies were simulated for each proliferation rate. b, Diagrams of cell state transition probabilities used for benchmarking PATHpro (Supplementary Note). c, Comparison of transition probability inferences between PATHpro and SSE-MLE for systems shown in b, when the true proliferation rates are known. Inference error is the Euclidean distance between inferred and true transition probabilities. d, Comparison of transition probability and proliferation rate inferences for networks shown in b when proliferation rates are not known. Inference error is measured as the Euclidean distance between true/simulated and inferred transition probabilities or proliferation rates and values reported are rounded to the nearest thousandth. e, Comparison of PATHpro and SSE-MLE run times for the inferences shown in d, with the y-axis is scaled by log₁₀(seconds). For all box plots, boxes represent the interquartile range (IQR); the center line represents the median; minima and maxima shown represent 1.5⋅IQR. For violin plots, white lines correspond to the median, and dashed lines represent the mean. Box and violin plots represent 100 simulated phylogeny replicates each.