Fig. 5

Clonality inference on the three-dimensional, spatial tumour model. To test the robustness of our results, we repeated the clonality inference process in the previously published spatial cancer model of Waclaw et al. [32], which is very different from our model. Small biopsies (green; one biopsy = 1 node) have a much greater probability to correctly classify clonal mutations than large biopsies (purple; one biopsy =8% of the total tumour size) if we include all mutations present in each sample ε=0.0. As we increase the threshold of the mutation frequency ε (middle panel ε=0.3, bottom panel ε=0.8), the accuracy of larger samples is increasing, and goes beyond single-cell samples and model predictions, same as in our simpler model. model. Number of tumours = 300 with maximum size of 5·10⁶ cells were simulated with a death rate of d=0.8. Mutation rate γ=0.02 mutations per division