Stochastic inference and control of complex biological networks

活细胞编码由相互连接的生物分子组成的复杂而动态的网络，这些网络协调着不同的生命过程。与电路不同，这些网络由生化物种(基因、蛋白质、RNA等)组成。它们通过化学反应相互作用和调节。全面了解生物分子网络如何为细胞决策处理信息将对人类健康产生巨大的长期影响。例如，生物分子网络的系统级破译将从根本上改变我们对驱动疾病状态的异常调控的知识，并将导致寻找考虑网络动态的生物标记物和药物靶标的新方法。此外，使用系统方法设计和重新布线网络将为不同的应用打开大门，例如生物燃料和治疗的生产。为了促进这些转化，本项目旨在建立可扩展的数学工具来对生物网络进行建模、分析、推理和控制。由于生化过程固有的概率性质，单个细胞内的测量揭示了生物网络具有丰富的随机动力学。智能的优点在于通过随机混合系统(SHS)框架对生物分子网络进行建模。该项目结合了控制论、动力系统和随机过程的工具，将开发易于计算的方法来分析复杂生物分子系统的确定性和随机动力学。与传统的蒙特卡罗模拟方法相比，这些方法将提高按数量级预测网络动态的计算效率。分析工具将用于探索生物分子系统中反馈和前馈环路的设计，这些环路允许系统地操纵网络活动，例如，控制嵌入更大网络中的特定蛋白质水平的波动。预测随机动态的准确方法也激发了一个有趣的推理问题，即从测量的网络组件的联合波动中学习底层网络体系结构。随着时间的推移，单细胞技术的进步使精确定量测量单个细胞内的蛋白质拷贝数成为可能。受此类测量可用性增加的推动，该项目将建立推断方法，以根据蛋白质水平的时间序列测量改进网络相互作用的表征和逆向工程。在更广泛的影响方面，项目工具将与生命科学实验研究人员密切合作，应用于研究不同的生物途径。其中包括几个医学上重要系统的调控网络，如人类免疫缺陷病毒(HIV)的细胞命运调控，以及癌细胞的非遗传耐药。最后，将为学生和来自当地行业的专业人员开发将系统和控制概念应用于生物网络的课程。

Living cells encode complex and dynamic networks of interconnected biomolecular components that orchestrate diverse life processes. Unlike electrical circuits, these networks consist of biochemical species (genes, proteins, RNAs, etc.) that interact and regulate each other via chemical reactions. A holistic understanding of how biomolecular networks process information for cellular decision-making will have tremendous long-term impact on human health. For example, a systems-level deciphering of biomolecular networks will fundamentally transform our knowledge of aberrant regulation driving diseased states, and will lead to novel ways of finding biomarkers and drug targets that take the network dynamics into account. Moreover, designing and rewiring of networks using systems approaches will open doors for different applications, such as, production of biofuels and therapeutics. To facilitate these transformations, this project aims to build scalable mathematical tools for modeling, analysis, inference and control of biological networks.Measurements inside individual cells reveal biological networks with rich stochastic dynamics, owing to the inherent probabilistic nature of biochemical processes. The intellectual merit lies in modeling of biomolecular networks via a Stochastic Hybrid Systems (SHS) framework. Combining tools from control theory, dynamical systems and random processes, the project will develop computationally tractable methods for analyzing deterministic and stochastic dynamics of complex biomolecular systems. These methods will improve computational efficiency of predicting network dynamics by orders of magnitude as compared to traditionally used Monte Carlo simulation techniques. Analysis tools will be used to explore designs of feedback and feedforward loops in biomolecular systems that allow for systematic manipulation of network activity, for example, controlling fluctuations in the level of a specific protein embedded in a larger network. Accurate methods for predicting stochastic dynamics also motivate an intriguing inference problem of learning about the underlying network architecture from measured joint fluctuations in the network components. Advances in single-cell technologies enable precise quantitative measurements of protein copy numbers inside individual cells over time. Motivated by increasing availability of such measurements, the project will build inference methods for improved characterization and reverse engineering of network interactions from time-series measurements of protein levels. In terms of broader impact, the project tools will be applied to study diverse biological pathways in close collaborations with experimental researchers in the life sciences. These include regulatory networks underlying several medically important systems, such as, cell-fate regulation in the human immunodeficiency virus (HIV), and nongenetic drug resistance in cancer cells. Finally, courses applying systems and control concepts to biological networks will be developed for both students and professionals from local industry.