Multi-objective identification of synthetic circuits stochastic models using flow flcytometry data

摘要

Synthetic biology use mathematical models of biological circuits to predict the behavior of the designed synthetic devices, but also to help in the design of the circuit and for the selection of their biological components. Estimation of these models parameters remains a demanding problem that has been addressed by optimization of a weighted combination of different prediction errors, thus obtaining only one solution. This single-objective approach can be inadequate when trying to incorporate different kinds of experiments or to identify parameters for an ensemble of biological circuit models and even more when dealing with stochastic models and flow cytometry data. Stochasticity in biological systems, often referred to as gene expression noise, is ubiquitous and needs to be taken into account when modeling a biological system. Here we present a methodology based on multi-objective optimization to perform parameter estimation in stochastic models using flow citometry data. It uses a global multi-objective evolutionary algorithm and a multi-criteria decision making strategy to select the most suitable solutions. We obtain an approximation to the Pareto set that corresponds to the model parameters better fitting the experimental data. Then, the Pareto set is clustered according to the different experimental cases, allowing to analyze the sensitivity of model parameters. We show the methodology applicability through the case study of a genetic circuit which controls noisy protein expression in a cell population.