Fig. 4

Analysis of traveling wave patterns. Based on average frequencies over 50 replicates, four statistics were measured for different fitness configurations: the number of driver waves within an superdriver evolution, width of waves, the difference in height between consecutive waves, and the number of generations between consecutive waves. We compared low superdriver selection (c ∊ {1.1, 1.3, 1.5}) and high superdriver selection (c ∊ {2.6, 2.8, 3.0}). Furthermore, we compared results between k ∊ {0, …, 5} superdriver and ℓ ∊ {0, …, 10} driver mutations. For all results, we fixed driver selection at s = 0.01. a Higher c leads to fewer driver waves conditioned on k and b all waves that span at least 500 generations tend to exist for fewer generations with less variance. c The difference in the maximum frequency of subsequent driver waves conditioned on k is lower for higher values of c; and d with higher c and conditioned on l, less generations lie between subsequent peaks of superdriver waves, suggesting that the maximum frequency of superdriver clones is reached in less generations