Agent-based simulations of lung tumour evolution suggest that ongoing cell competition drives realistic clonal expansions

Computational simulations of tumour evolution are increasingly used to infer the rules underlying cancer growth, with the goal of one day recommending tailored treatments. To make reliable inferences, such models must be able to reflect the properties of real tumours. Recent work has shown that lung tumours undergo frequent and late subclonal expansions, which are associated with poor prognosis. This paper tests three candidate simulations of three-dimensional tumour growth, which make different assumptions about the nature of competition between cells, for their ability to replicate these late expansions. Only a model which assumes stringent competition between tumour cells for existing tissue space, after the tumour has reached a fixed size, can produce multiregion sequencing data realistic to lung tumours. This work also assesses the influence of these assumptions on the inferred selection strength of individual tumours in a large cohort of lung cancers, and finds that inferences from a well-tailored model produce much higher estimates than poorly-tailored models of the effects of driver mutations on cell fitness.