摘要：Multi-objective optimisation is a valuable tool for tuning dynamical systems when simultaneous optimisation performance objectives are in conflict. When the goal is tuning the parameters of a synthetic biology device, mismatch between the model implementedin silico-a more or less coarse simplification of the real system- and the actualin vivoimplementation might lead to a disagreement between thein silicoandin vivodesign objectives for a given solution from the Pareto front. Here, we propose an iterative closed-loop multi-objective optimisation approach where the new information provided by the difference between thein silicoPareto front and itsin vivoimplementation is used to improve the parametric model. This aims to minimise the discrepancies betweenin silicoandin vivoperformance objectives while preserving the trade-off order among solutions. As a proof-of-concept we consider the problem of tuning a synthetic gene circuit used as feedforward-feedback controller for the expression of a protein of interest. We use an extended parametric model of the gene synthetic circuit to represent thein vivoset up and a simplified one for thein silicoone.
