Breaking records in the evolutionary race

We explore some aspects of the relationship between biological evolution processes and the mathematical theory of records. For Eigenʹs quasispecies model with an uncorrelated fitness landscape, we show that the evolutionary trajectories traced out by a population initially localized at a randomly chosen point in sequence space can be described in close analogy to record dynamics, with two complications. First, the increasing number of genotypes that become available with increasing distance from the starting point implies that fitness records are more frequent than for the standard case of independent, identically distributed random variables. Second, fitness records can be bypassed, which strongly reduces the number of genotypes that take part in an evolutionary trajectory. For exponential and Gaussian fitness distributions, this number scales with sequence length N as , and it is of order unity for distributions with a power law tail. This is in strong contrast to the number of records, which is of order N for any fitness distribution.