evosim: fast and scalable stochastic simulations of evolutionary dynamics

Abstract

The simulation of clonal dynamics with branching processes can provide valuable insights into disease progression and treatment optimization, but exact simulation of branching processes via the Stochastic Simulation Algorithm (SSA) is computationally prohibitive at the large population sizes associated with therapeutically-relevant scenarios. evosim is a versatile and flexible Python implementation of a fast and unbiased tau-leaping algorithm for the simulation of birth-death-mutation branching processes that is scalable to any population size. Package functionalities support the incorporation and tracking of a sequence of evolutionary changes such as therapeutic interventions as well as the analysis of population diversity. We show that runtimes scale logarithmically with population size, by contrast to the linear scaling of the SSA, and simulations exhibit strong agreement with SSA simulation results. These findings are also supported by mathematical results (Supplementary information). AvailabilityPackage, documentation, and tutorials / usage examples are available on GitHub (https://github.com/daliten/evosim). Mathematical details of the algorithm and the pseudocode are provided in the included Supplementary information.