Identification and control of gene networks in living cells

System identification is a branch of control engineering aimed at developing computational approaches to derive, from measurement data, a quantitative dynamical model of a physical system able to predict its future behaviour. There is a long tradition in the successful application of system identification approaches to Medicine and Physiology, however, in Molecular Biology, only few attempts have been made to infer a quantitative model of gene regulation due to experimental limitations of current techniques. Indeed, whereas in Engineering it is now common to measure thousands of time-points at a desired sampling rate for a physical system to be modelled, this has been very difficult in Biology, where time-series data consist of very few samples. In order to overcome the current limitations, I devised an experimental platform based on a microfluidic device, a time-lapse microscopy apparatus and, a set of automated syringes all controlled by a computer, that allows to provide a time varying concentration of any molecule of interest (input) to a population of cells, and to measure the single-cell response in the form of the fluorescence level of a reporter protein, at a sufficiently high sampling rate, thus making it possible to evaluate the dynamics of the process of interest. I tested the experimental platform to implement and compare different linear and nonlinear system identification approaches to a transcriptional network in the yeast S. cerevisiae. The results I obtained confirm that the experimental system identification platform I developed can successfully be used to infer quantitative models of a eukaryotic promoter in a rapid and efficient manner. Moreover I have used the same experimental set up for the study and the it in-vivo implementation of novel feedback control strategies meant to precisely regulate the level of expression of a protein from the GAL1 endogenous promoter and from a complex synthetic transcriptional network in yeast cells. The proposed effective control approach, allows to generate custom time profiles of a desired protein, and it can be exploited to study trafficking or signalling pathways and the endogenous control mechanisms of a cell.