<i>In silico</i> simulation under fluctuating environments and adaptation time for genotypes that are robust and evolvable.

<p>(a): Monostable genotypes corresponding to the population mean of parameter values arising from different generations of a single <i>in silico</i> simulation under fluctuating environments, superimposed onto the G-P mapping of circuit I (grey backdrop). Genotypes derived from each generation are color-coded according to generation number, with red indicating the oldest generations. For both the G-P mapping and the <i>in silico</i> simulations, evolvability was computed using parameter perturbations of ±20% (cf. Figure F in <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0153295#pone.0153295.s001" target="_blank">S1 File</a>)), and robustness was calculated using the sensitivity-based measure. See also Figure G in <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0153295#pone.0153295.s001" target="_blank">S1 File</a>, which shows the results obtained for other runs of the evolutionary simulation algorithm. (b): Box plots showing adaptation time for genotypes exhibiting high evolvability (<i>H</i>) and low evolvability (<i>L</i>) under two selection scenarios: selection for a 10% reduction (blue) and for a 10% increase (red) in steady state expression levels. For circuit I, high evolvability (<i>H</i>) was <i>E</i> > 0.7 and low evolvability (<i>L</i>) was <i>E</i> < 0.1 on G-P mappings using robustness and evolvability measures computed from 10% parameter perturbations (cf. <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0153295#pone.0153295.g002" target="_blank">Fig 2c</a>). For circuit II, genotypes from <i>H</i> had <i>E</i> > 1.2 and genotypes from <i>L</i> had <i>E</i> < 0.6 based upon the same measures (cf. <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0153295#pone.0153295.g002" target="_blank">Fig 2f</a>). For both circuits, 11000 genotypes were picked from each respective region in the G-P mapping. For both circuits, adaptation times for the two regions differed significantly (<i>p</i> < 0.01, two-sample Kolmogorov-Smirnov test, n = 11000).