Fig. 1.

Phenotypic interference. Phenotypic interference is an evolutionary mode of multiple quantitative traits under genetic linkage with the following features: a Local evolution. Individual traits G change by mutations (red arrows) with mean square selection coefficient s² set by the slope of the fitness landscape f(G). In housekeeping equilibrium, the traits have detailed balance between beneficial and deleterious substitutions³⁰. b Global evolution proceeds in a fitness wave, i.e., multiple genetic and phenotypic variants co-exist in a population^{5,7–9,11–13}. The wave is fueled by new beneficial mutation at its tip (red arrows). It has fitness variance σ² and a total fitness span $\widehat{\sigma }=\sqrt{c}\sigma$ with a factor c depending weakly on the population size (see Methods section). By Fisher’s theorem, these quantities determine the coalescence rate $\tilde{\sigma }={\sigma }^2∕\widehat{\sigma }=\sigma ∕\sqrt{c}$ (see Methods section). Its inverse is the coalescence time (i.e., the average time to the most recent common ancestor of two individuals), which is proportional to the effective population size, ${\tilde{\sigma }}^{-1}=2N_e$ (red bar)^13,14. In housekeeping equilibrium, the selective advance (light gray) is offset by deleterious mutations, i.e., there is no net (local or global) change in fitness^12,77 (dark gray). Feedback between global and local evolution: the global coalescence rate sets selection on individual traits, $s^2=4{\tilde{\sigma }}^2$ (Eq. 3). The wave complexity, measured by the number of simultaneously segregating beneficial trait mutations destined for fixation, is of order $2\widehat{\sigma }∕s=c$ (Eq. 21). c Superlinear load. The genetic load $\mathcal{L}$ depends quadratically on the number of genes, g, over a broad range $g_0\lesssim g\lesssim g_{\mathrm{m}}$ (red line, see text). This load is substantially higher than the load under asexual evolution with discrete gene fitness effects (blue) and under of recombination (brown)