a, Schematic representation of mouse model. b, Tumour growth kinetics (mm3) of individual tumours (yellow), and corrected mean obtained from extrapolating the missing time-points using the model prediction (black curve). c, Confocal images of melanomas exhibiting Confetti expression at different relative growths upon TAM administration (0.5 mg). Representative images from at least 8 mice per group. d, Representative reconstruction of 3D clones. Independent clones appear in different colours. A single z-slice of the corresponding dataset is shown at the bottom of the plot. e. Dot plot showing the distribution of clone sizes, by sample and by channel obtained from 5 cleared tumours (see Supplementary Table 5). f. Cumulative distribution, showing the probability of observing a clone larger than a given size. The corresponding bi-exponential fits (dashed line, R2=0.96, standard error of the fit S=0.03) and exponential decay of the tail (black line) are shown. g, Schematics of the minimal hierarchical model. The parameters ΔS and ΔP correspond to the expansion rates of the S and P populations, respectively. Although the data could accommodate a “line” of fit parameters, fixing the relative transition rate of stem cells into the progenitor cell compartment, r=0.75, the model shows an excellent fit to the data with ΔS~0.09 and ΔP~0.04 and a stem fraction fS~0.21 or 21% (see Supplementary Note). h, Upper panel, experimental procedure for dual pulse labelling assay using EdU and BrdU. Lower panel, image of a double BrdU (green) and EdU (red) immunostaining. Representative image from n=5 tumors. i, Pie-chart illustrating the percentage of single and double labelled cells. Quantification was assessed from 5 different mice and 4 technical replicates per mouse. Schematic in a, g and h made with Biorender.com.