Table 1.
Parameters and variables.
Fig 1.
Population dynamics of a single parameter set.
A) Trait evolution after an optimum shift and B) allele frequency dynamics during adaptation from a single replicate. The vertical line shows when the new trait optimum was reached and line colors denote effect sizes. Time is shown on a log scale. C) The phenotypic distribution and D) site frequency spectra of segregating mutations (black) and neutral expectation (red) from 100 independent replicates. Panels show different generations including equilibrium prior to adaptation (0), during adaptation (0.005), just before the new optimum is reached (0.01), after the new optimum has been reached (0.02), and the final generation (0.1). All results are from a simulated population with constant population size, σm = 0.05, and VS = 1.
Fig 2.
A) Speed of fixation of selective sweep mutations. B) Dynamics of fixations that occur after the new optimum was reached. C) Speed of fixation of sweeps from standing variation compared to their initial frequency. D) The generation at which sweeps from standing variation fix. All results are from a simulated population with constant population size, σm = 0.05, and VS = 1, and time is shown on a log scale.
Fig 3.
Relative parameter importance.
Relative parameter importance inferred for four parameter categories. 1) Adaptation: parameters describing adaptation speed and potential for future adaptation, 2) Fixations: summary statistics for mutations that were fixed during trait adaptation, and 3) Segregating sites: descriptors of alleles polymorphic in the final generation of the simulations. Top rows indicate prediction accuracy as calculated by 10-fold cross validation and NRMSE. Each bar is the result of an independent random forest learning and each color represents the relative importance of the simulation input parameters (see Methods and S1 Table. for summary statistics).
Fig 4.
Summary of trait adaptation and selective sweeps.
A) Time to reach new trait optimum B) Rate of change in phenotype C) Genetic variance after 0.1 × Nanc generations. D) Total number of selective sweeps, separated by type of sweeps. E) Proportion of sweeps compared to all fixations F) Proportion of sweeps from standing variation. Boxes are split by major parameter importance as identified by our random forest model. Points in A-C and E-F show the values of each of 1,200 parameter sets and are colored according to bottleneck size (Darker color indicate stronger bottleneck, see legend in A). Interactive plots are available at https://mgstetter.shinyapps.io/quantgensimAPP/.
Fig 5.
Genetic architecture in final population.
A) Genetic architecture matrices for two parameter combinations (maize models, see Methods) differing in effect size of new mutations and strength of stabilizing selection. Effect size bins are centered around zero with negative effect size quantiles on the left and positive quantiles on the right of the central bin. Shown is the correlation coefficient between the genetic architectures. B) Pairwise correlation of genetic architecture of all comparisons of 1,200 parameter combinations. Subplots display the combination of final population size (log; 1, 3, 10) and effect size distribution (σm, 0.01, 0.05, 0.1, 0.3, 0.9) of incoming mutations. Each pixel displays a pairwise comparison between two of the 1,200 scenarios. C) Relative parameter importance for genetic architecture prediction.
Fig 6.
A) The evolution of trait value and B) genetic variance during adaptation to a new trait optimum for two traits under maize demography with no genetic background. Time in both figures is shown on a log scale, light shadows show standard deviations from the mean of 100 simulation replicates. Trait 1 (blue) has small effect mutations (σm = 0.01) and strong stabilizing selection (VS = 1). Trait 2 (red) has large effect new mutations (σm = 0.9) and weak stabilizing selection (VS = 50). Vertical lines denote the generation when 99% of the new trait optimum is reached. C) Proportion of selective sweeps. D) Sojourn time of sweeps from standing variation in Trait 1. E) Sojourn time of sweeps from standing variation in Trait 2. Scales in D and E are different due to strong divergence of effect size values.