Extended Data Fig. 3: PATH inferences and simulations of somatic evolution.
a, Same as Fig. 2b but for systems where heritability is not detectable for all cell states (at least one state phylogenetic autocorrelation z score ≤ 2) b, Comparing the state transition dynamic inference accuracy of PATH using phylogenetic correlations measured at different node depths. Transition inference error is measured as the Euclidean distance between simulated and inferred transition probabilities, and the number of possible states in a simulated system is shown on the x-axis. 1,000 phylogenies were simulated for each system and sampled at a rate of 10−2. c, Mean path length distance, phylogenetic correlation difference, and Robinson-Foulds distance between simulated true and reconstructed phylogenies (Supplementary Note). Phylogenies were simulated 1,000 times for each barcode length (x-axis). d, Expected pendant edge lengths for a sampled somatic evolutionary process, as a function of birth, death, and sampling rates (Supplementary Note). e, Transition inference error (Euclidean distance between inferred and true transition probabilities) using PATH or MLE for 3, 4, or 5 cell states in a phylogeny composed of either 100, 500, or 1,000 cells, representing a sample of 10−6 or 10−3 of the total population. Each parameter combination was simulated 1,000 times. f, Run times corresponding to simulations sampled at a rate of 10−6 depicted in e. Box plots represent median, bottom and upper quartiles; whiskers correspond to 1.5 times interquartile range.
