Figure 3.

Distribution of randomly sampled cells on yield–risk space. A heterogeneous culture of cells is composed of a mixture of wild type and various mutant cells. (A) At the initial stage of culturing, random sampling of cells and mapping onto the yield–risk space may result in a broad distribution for the yield and risk of each cell. (B) Culturing this population under the stationary condition for multiple passages may result in evolution of the culture toward a high-yield genotype. Each circle represents randomly sampled cells at this time point, and not the same cell sampled in (A), because multiple passages have occurred. (C) Further passages under the stationary condition result in a distribution strongly favoring high-yield individuals. If there is a trade-off between robustness and performance, hence an inverse of risk and yield holds, distribution will be constrained within a certain envelope represented by the efficient frontier. Artificial evolution experiments with random sampling for yield–risk space mapping test the conjecture of robustness–performance trade-off at the efficient frontier. (D) If the robustness–performance trade-off holds at any time, even without being at the efficient frontier, the center of gravity of all randomly sampled points after passages will simply shift toward the upper right. This contrasts to the case that the trade-off holds only at the efficient frontier. This difference can be experimentally verified.