in silico生物分子网络动力学参数高速与高精度自动化估计的研究

■ Background: Simulation studies on biological systems promises a deep understanding of complex cellular mechanisms by investigating dynamic features. How to determinate kinetic parameters which govern the results of a simulation model is one of the most critical problem of current research in systems biology. The complexity of biological systems and the progress of experimental techniques severely limit the acquirement of appropriate kinetic parameters. Meanwhile, existing methods also have some issues on estimating kinetic parameters. Data assimilation (DA) approach enables users to handle parameter estimation in the context of statistical inferences. However, it highly depends on providing successive time points of time-series data and costs massive computational resource for the estimation, in some cases, it will be completely impossible to estimate parameters. .■ Method: We therefore propose a novel computational framework DAMC to automatically estimate kinetic parameters of biological networks by introducing our new generic Probabilistic Linear-time Temporal Logic (gPLTL) based model checking (MC) to data assimilation. We employ a statistical method -- particle filter, often applied to DA for our estimation procedure. We evaluate the availability and practicality of proposed method by a high-level Petri net model underlying circadian rhythm in mouse as an example. .■ Result and Conclusion: The contributions of this research are as follows: (i) DAMC provides a reliable theoretical basis for the analysis of large-scale biological networks by means of supercomputers; (ii) it is a great help in improving the efficiency and accuracy of conventional hand tuning method and reducing the search space of the particle filter; We can expect to unravel insightful biological knowledge and to forecast reactions of the biological systems to the drugs and chemical compounds in more plausible ways based on the combined information from the both simulation results and experimental data.

计算机仿真在分析生物系统的动力学特性，揭示生物系统的作用机理中发挥着重要作用。如何确定影响仿真结果的反应动力学参数是目前系统生物学面临的关键科学问题之一。由于生物系统本身的复杂性与实验测量技术的局限性，使我们无法确切地获得动力学参数，目前主要的计算估计方法也存在一定弊端。本项目是基于数据同化只局限在十几个参数的估计，并对时序列数据及计算机配置有较高依赖的研究背景下，拟在高阶Petri网理论基础上，提出新的一般概率线性时态逻辑gPLTL的模型检测理论并将其整合至数据同化方法中，建立高速与高精度仿真模型的动力学参数自动化估计框架DAMC，并在小鼠昼夜节律调控网络上评价DAMC的有效性和实用性。本研究可摆脱传统手动设置参数的方式，减少主观人为因素干扰。不仅能大幅度提高计算效率减少单次参数估计过程中使用的计算空间，并且对于将来展开使用高性能计算机对大规模生物系统网络进行分析提供有力的理论基础。

计算机建模仿真在分析生物系统的动力学特性，揭示生物系统的作用机理中发挥着重要作用。如何构建可靠及正确的生物网络动力学模型是目前系统生物学面临的关键科学问题之一。由于生物系统本身的复杂性与实验测量技术的局限性，使我们无法确切地获得动力学参数，目前主要的估计方法也存在一定弊端。本项目提出并建立了高速与高精度仿真模型的动力学参数自动估计框架DAMC来构建可靠的大规模生物网络模型。DAMC算法摆脱传统手动设置参数的方式，减少主观人为因素干扰；同时减少对时序观察数据及计算机高配置的依赖。我们通过小鼠昼夜节律调控网络的模型验证了DAMC算法的有效性及实用性。该计算框架的建立对于将来展开使用高性能计算机对大规模生物系统网络进行分析提供有力的理论基础。项目资助期间已发表论文3篇，出版书籍1本，受邀参加了1个国际会议和2个国内会议，口头汇报了关于DAMC研究思路及现阶段工作结果。