The "New Synthesis"

Abstract

When Mendel's work was rediscovered in 1900, and extended to establish classical genetics, it was initially seen in opposition to Darwin's theory of evolution by natural selection on continuous variation, as represented by the biometric research program that was the foundation of quantitative genetics. As Fisher, Haldane, and Wright established a century ago, Mendelian inheritance is exactly what is needed for natural selection to work efficiently. Yet, the synthesis remains unfinished. We do not understand why sexual reproduction and a fair meiosis predominate in eukaryotes, or how far these are responsible for their diversity and complexity. Moreover, although quantitative geneticists have long known that adaptive variation is highly polygenic, and that this is essential for efficient selection, this is only now becoming appreciated by molecular biologists-and we still do not have a good framework for understanding polygenic variation or diffuse function.

Keywords: infinitesimal model; polygenic adaptation; quantitative genetics.

Fig. 1.

Response of a trait, and of the underlying allele frequencies, to a change in optimum from z_opt = 0 to 10 at time zero. (A) Trait mean (Left) and variance over time (Right). The mean responds rapidly, reaching the new optimum within 50 generations. The variance increases sharply, as + alleles increase in frequency, and then gradually returns to its original value over thousands of generations, as stabilizing selection acts to reduce the variance. (B) Frequencies of the + allele, plotted for the 95 loci with effect $\gamma >0.3$; the plot is on a logit scale, so as to expand frequencies near to fixation. Allele frequencies increase sharply as the mean moves to the new optimum, but then slowly return toward fixation, as stabilizing selection acts to reduce heterozygosity. Ultimately, only seven of these alleles with $\gamma >0.2$ substitute, contributing a change in mean of 3.82 out of the total of 10. (C) The contribution of each locus to the change in trait mean, $\gamma \text{Δp}$, plotted against its effect. By 100 generations after the change in optimum (Left), only the largest-effect alleles have shifted substantially; alleles with effect $\gamma >0.2$ account for a change in mean of 4.1 out of a total 9.6, with seven loci contributing 2.0. Ultimately (Right), large-effect loci have either fixed (upper diagonal line, $\text{Δp}$ = 1) or returned to low frequency ($\text{Δp}$ ≈ 0); somewhat less than half the ultimate change in mean (3.8 out of ∼10) is contributed by substitutions of large-effect alleles, the remainder being due to small shifts in frequency of weakly selected alleles. There are 1,000 biallelic loci, with additive effects $\gamma$ drawn from an exponential distribution with mean 0.12. Mutation is symmetric, at rate $\mu$ = 0.000025; fitness is $\sim \text{exp}[{\left(z-z_{\text{opt}}\right)}^2/2V_s].$, with V_s = 20. There are n = 10⁴ haploid individuals. The simulation is run for 10⁴ generations, to reach equilibrium, when the mean is close to the optimum, and the genetic variance maintained in a balance between mutation, drift, and stabilizing selection is V_g = 0.556; this is close to the prediction from the diffusion approximation [red line in A, Right (ref. , equations 6 and 7)]. See SI Appendix for details.