Fig. 2

Comparison of clonality inferences in structured and unstructured models of tumours with a different proportion of the largest sub-clone. a The probability to correctly identify set of truly clonal mutations with n tumour samples in our model. In tumours where the size of the largest sub-clone f is small (f=0.1) the probability to correctly identify truly clonal mutations is already sufficiently (> 98%) high with two samples. In balanced tumours with f=0.5, five samples give the same probability. b The quality of our clonality estimation in dependence of the proportion of the first sub-clone. Lines represent solutions of the mathematical model, while dots represent results from clonality inferences of spatial computer simulations. A number of randomly distributed single-cell samples n (n=2 shown in blue, n=5 shown in red and n=10 shown in green) was taken from each simulated tumour and clonality of present mutations was estimated. Each single dot represents the proportion of correct estimations for one tumour by sampling n tumour samples after 10 000 repetitions. Results from simulations are in agreement with model predictions for the full range of f