Fig. 9: Exploring the space of possible models of population control to find the best system.

a Schematic of all possible components for single strain engineering. The engineered strain can produce an intercellular toxin and its antitoxin and an intracellular toxin and its antitoxin. All of these can be positively or negatively regulated by a self-produced quorum molecule, allowing density-dependent control of the engineered cell’s actions. b Once the model space has been defined, we explore it using ABC SMC to determine which models with which parameters are capable of producing stable coexisting populations. c The ordered model marginal posterior probabilities. The long tail of extinct models has been trimmed. The boxes for each model are coloured according to the model’s complexity; models with a larger number of parameters are closer to the yellow and fewer parameters are close to blue. d The best system (i) requires control over the expression of four genes. Systems that do not require intracellular toxin or immunity perform well for their level of complexity (ii–iv). In fact, the simplest system able to produce stable co-existence is the system that we have developed (iv). In c, each box is centred at the mean, with the left and right edges showing the standard deviation, of three replicates, each containing 18,000 particles. Black points show the outcome of each replicate.