Fig. 2

The stochasticity of appearance of the first non-terminal arrest is generation-dependent and best described by a variable Bernoulli parameter with an exponential growth. A Grey bars: distribution of the occurrence of the first non-terminal arrests with generations. Dots: Fit of the distribution with p(j) following the logistic function shown in B (blue dots) or being constant (red dots). B Grey dots: experimental frequency of occurrence of the first non-terminal arrest; blue line: logistic fit of p(j), the probability of appearance of the first non-terminal arrest, calculated only for generations with more than 5 lineages left for computing. C Ordered generations of the first arrest from the experimental data (black line, n = 115 lineages) or simulations (N = 1000) based on Bernoulli distribution of first arrest with an exponentially-growing parameter with the blue shaded area representing the 95 % quantile. (D) p-values of χ² goodness-of-fit tests of X corresponding to Bernoulli trials with a non-constant p(j) fitted in (A) as a function of the threshold D. The best parameters of (2) are computed for each D. Red line represents p-value = 0.05. The hypothesis is not rejected for D between 180 and 270 min