Fig. 3: The impact of compensatory mutations on treatment failure is delayed.

a, b Treatment failure probabilities of experiments with different mutation rates for early (a, day 5, n = 6 independent colonies for each mutation rate) and late (b, day 9, n = 18 independent colonies for each mutation rate) hygromycin treatment (Eq. (3)) as measured via the number of resurgent growth domes. The fitness cost of all experiments was s = 0.013 ± 0.006 (Supplementary Fig. 2a, b). Mutation rates were either low (μ = 5.6 ± 3.5 × 10⁻⁶ μm⁻¹), medium (μ = 2.65 ± 0.25 × 10⁻⁴ μm⁻¹), or high (μ = 4.48 ± 0.34 × 10⁻⁴ μm⁻¹) (Supplementary Fig. 2c, d). Colors represent the fraction of compensated (red) and uncompensated (blue) growth domes. Numbers refer to the percentage of compensated clones. The difference between the low mutation rate control and experiments with compensatory mutations are non-significant for treatment after day 5 (a, p-values 0.58 and 0.22 for medium and high switching rates, respectively) but very significant for treatment after day 9 (b, p-values 4 × 10⁻²² and 7.5 × 10⁻³⁰), one-sided t-tests. Errors of P_fail indicate one SD of Poisson distribution. c, d Efficacy of evolutionary rescue with reference to low mutation experiment for treatment at day 5 or day 9, respectively. Error bars indicate propagated SDs. Source data are provided as a Source data file.