“…The empirical foundation of our method is merely a standard competition assay but one in which the superior type arises through native mutation rather than by experimental addition of known mutants ( 6 , 10 , 33 , 43 ). Our maximum likelihood estimation is suited to the simplicity of our underlying model, but the estimation could be expanded to considerable model complexity by using Approximate Bayesian Computation (ABC), albeit that the ABC approach may involve considerably more empirical analysis and computational trial and error (e.g., 2 ). We also acknowledge that similar but alternative approaches have been developed, specifically tailored to the problem of plasmid maintenance ( 4 , 5 , 22 , 26 ).…”