Model Reduction and Multi-objective Identification of a Feedback Synthetic Gene Circuit

TitleModel Reduction and Multi-objective Identification of a Feedback Synthetic Gene Circuit
Publication TypeJournal Article
Year of PublicationSubmitted
AuthorsBoada Y, Vignoni A, Picó J
JournalIEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY
Abstract

Identification of model parameters is an es- tablished problem in control systems technology. In sys- tems and synthetic biology kinetic (i.e. dynamic) semi- mechanistic models based on first principles are particu- larly important since they can explain and predict the func- tional behavior that emerges from the time-varying con- centrations in cellular components. However, these mod- els have high dimension and a large number of parame- ters. Therefore, systematic model reduction techniques are needed. The resulting reduced nonlinear models present incomplete parameter identifiability. Moreover, suitable pa- rameter identification methodologies are required to cope with the noisy and partially observed nature of data in biology. Thus biological systems identification still appears as an open problem having to deal with complex nonlinear dynamics, incomplete parameter identifiability, and scarce experimental data of different natures. Multi-objective opti- mization arises as a natural option in this scenario. Here, propose and apply a systematic methodology for model reduction of genetic circuits, and identification of parame- ters of the resulting reduced model using a multi-objective identification framework. We apply the methodology to a feedback loop synthetic genetic circuit designed to mini- mize gene expression noise. To enhance the information content of the measurable variables, we first identify an open loop version of the circuit using averaged time-course experimental data obtained from plate readers. Then, we use steady state stochastic distributions provided by flow cytometry to identify the remaining feedback gain. Our methodology gives good identification results for ensemble models and it is particularly useful for easily combining experimental flow cytometry data with experimental plate reader data.