APPLYING FEEDBACK CONTROL TO AUTOMATICALLY TRACK AND DESIGN COMPLEX DYNAMICS IN SYSTEMS AND SYNTHETIC BIOLOGY

数学建模在系统生物学和合成生物学中被广泛应用，以了解非线性生化现象(如基因表达的时间振荡)，并设计和验证工程基因电路。虽然生物化学模型得到了广泛的应用，但在其推导和相关参数识别方面都具有挑战性。参数和模型的不确定性可以显著改变非线性分子动力学行为的模型预测的可靠性。这反过来又挑战了使用模型来设计有效的癌症治疗方法，这些治疗方法针对多个信号通路，以避免耐药性和适应性。最近，在合成生物学中，控制工程的原理已被应用于控制活细胞中的基因表达。所谓的“外部反馈控制”利用微流控/显微镜平台来培养和干扰动态活细胞，同时通过荧光报告和分割算法来监测基因的表达。外反馈控制在哺乳动物细胞中的应用是非常新的[1]，到目前为止，只提供了概念证明，继电器和模型预测控制策略可以用来调节外源和内源基因的表达。本项目旨在改进和扩展哺乳动物细胞外反馈控制的现有能力，并探索其在系统和合成生物学中的应用。我们旨在实现的改进方法包括改进分割算法以包含深度学习方法，采用无模型策略(如自适应模型预测控制)，能够应对基因表达调控的随机性和系统数学模型中的不确定性。在开发这些工具的同时，我们将测试外部反馈控制策略的可用性，以达到以下目的：1)自动从实验中映射内源和外源基因调控网络的非线性动态特征，应用基于控制的延续，迄今其应用仅限于机电系统；2)设计优越的组合疗法，以克服肺癌细胞系中出现的耐药性；该项目的这一部分将与阿斯利康合作开发，目前与EPSRC的主要主管(Dr Marucci)合作。这一跨学科项目的成果将影响系统与合成生物学以及控制工程社区的研究人员。美国计算机学会合成生物学2018

Mathematical modelling is widely used in System and Synthetic Biology to understand nonlinear biochemical phenomena (e.g. gene expression temporal oscillations), and to design and validate engineered gene circuits. Although widely used, biochemical models can be challenging in both their derivation and associated parameter identification. Parameter and model uncertainty can significantly alter the reliability of model predictions of nonlinear molecular dynamic behaviours. This, in turns, challenges the use of models to design effective cancer treatments which target multiple signaling pathways to avoid resistance and adaptation.Recently, in Synthetic Biology, principles from Control Engineering have been applied to steer gene expression in living cells. The so called "external feedback control" exploits microfluidics/microscopy platforms to culture and perturb dynamically living cells, while monitoring gene expression by means of fluorescent reporters and segmentation algorithms. The application of external feedback control in mammalian cell is very recent [1] and has, so far, only provided the proof-of-concept that Relay and Model Predictive Control strategies can be used to regulate exogenous and endogenous gene expression.This project aims both at improving and extending current capabilities of mammalian cell external feedback control, and at exploring its application to address important open challenges in Systems and Synthetic Biology.The improved methodologies we aim at implementing include both refining the segmentation algorithms to encompass Deep-Learning approaches, and adopting model-free strategies (e.g. Adaptive Model Predictive Control) able to cope with the stochastic nature of gene expression regulation and uncertainties in system mathematical models. While developing these tools, we will test the usability of external feedback control strategies to:1) automatically map from experiments nonlinear dynamic features of endogenous and exogenous gene regulatory networks applying Control-based continuation, whose application so far has been limited to electro-mechanical systems; 2) design superior combination therapies to overcome the emergence of drug resistance in lung cancer cell lines; this part of the project will be developed in collaboration with AstraZeneca, currently partnering with the main supervisor (Dr Marucci) in an EPSRC Fellowship.Outcomes of this interdisciplinary project will impact researchers across the Systems and Synthetic Biology, and Control Engineering communities.[1] Postiglione et al. ACS Synthetic Biology 2018