Dynamics of clonal diversity (inverse Simpson index D) in 20 stochastic simulations of a non-spatial model. Black curves correspond to the individual simulations illustrated in subsequent columns. These particular simulations are those with indices D and n closest to the medians of sets of 100 replicates. Second column: Muller plots of clonal dynamics over time. Colours represent clones with distinct combinations of driver mutations (the original clone is grey-brown; subsequent clones are coloured using a recycled palette of 26 colours). Descendant clones are shown emerging from inside their parents. Third column: Final clone proportions (for the non-spatial model) or spatial arrangement (for spatial models). For spatial models, each pixel corresponds to a patch of cells, corresponding to a tumour gland, coloured according to the most abundant clone within the patch. Fourth column: Driver phylogenetic trees. Node size corresponds to clone population size at the final time point and the founding clone is coloured red. Only clones whose descendants represent at least 1% of the final population are shown. Final column: Evolutionary indices D and n at the final time point. Black points correspond to the individual simulations illustrated in previous columns. a, A model of tumour growth via gland fission (8,192 cells per gland), in which cells can acquire driver mutations that increase their contribution to the gland fission rate (with an average effect size of 50%), in addition to drivers that increase the cell division rate. b, A model in which tumour cells invade normal tissue but do not disperse within the tumour bulk (512 cells per gland). c, A boundary-growth model of a non-glandular tumour in which cells invade neighbouring sites within the tumour. d, A model in which tumour cells invade normal tissue at the tumour boundary only (2,048 cells per gland). e, A model of tumour growth via gland fission (2,048 cells per gland). Other parameter values are listed in Supplementary Table 4.