Fig. 1

a Pipeline for data analysis with the mobster R package for subclonal deconvolution. The package can be used to infer the clonal composition of a tumour bulk DNA biopsy. The statistical method integrates machine learning and evolutionary theory to detect subclones that have expanded due to positive selection, while modelling intra-clone neutral evolution with distributions predicted by population genetics. mobster also computes clone-specific evolutionary parameters (e.g., mutation rates, selection coefficients), dN/dS statistics and clone trees. b Example tumour clonal expansion simulated with the stochastic branching process implemented in the TEMULATOR (https://t-heide.github.io/TEMULATOR/) package. After \(t=13\) cell doublings a subclonal driver triggers a clonal expansion that sweeps in the ancestral population (left). Using mobster from a bulk sample collected at \(t=17\) we can retrieve the tumour architecture, the clone tree and the evolutionary parameters of the simulated tumour (right)