Hardening Software for Rule-based Modeling

项目摘要/摘要 基于规则的建模方法，它基于化学动力学和扩散原理以及 借助不断扩大的复杂软件工具(例如BioNetGen/NFsim)，提供SPE- 研究多位点信号蛋白相互作用动力学的社会优势。基于规则的模式- ELS可以捕捉类聚合反应和多位翻译后修饰的影响 时间尺度从几秒到几小时，同时结合了分子结构施加的限制。此外， 使用基于规则的模型制定方法，可以构建和分析更大、更复杂的模型。 与传统的建模方法相比，用于细胞调控系统的大量模型是因为运筹帷幄。 使用生物分子的形式规则在高抽象水平上简明地表示系统的机会 互动。通常可以对规则进行处理，以自动派生传统的模型形式，如耦合的 常微分方程组(ODE)。然而，当规则所隐含的系统状态空间为 非常大的规模，使用基于无网络算法的模拟引擎变得必要 随机模拟计算成本高，限制了模型分析的进行。此外，在这些电路中- 在实际应用中，参数辨识和不确定性量化(UQ)是极具挑战性的。 我们将通过提高模拟、拟合和UQ工具的效率以及利用 ING分布式计算资源。最近，我们提出了加速随机模拟的新算法。 拟合的并行元启发式优化方法工具箱及其马尔可夫实现 贝叶斯统一队列的链蒙特卡罗方法。这个名为PyBioNetFit(PyBNF)的工具箱利用 用于定义和共享模型的标准化格式(例如，核心SBML和BNGL)，并与var- 借条模拟器。在这里，我们建议开发用于加速网络的通用软件实现- 无工作(随机)模拟和重组基于规则的模型(即，优化规则以最小化 调整包含规则的方程式的数量)。我们还将为CVODE和CVODES提供一个新的接口 常微分方程组的数值积分、正演灵敏度分析和伴随灵敏度分析。此外，我们 我将对PyBNF的生物特性规范语言(BPSL)进行扩展，以使这一手段形式化 定性数据更具表现力。此外，我们还将添加基于梯度的优化方法和MCMC方法 并内置了对Smoldyn的支持，Smoldyn是(基于规则的)空间随机模型的模拟器。这些信息- 𝜀𝜀 验证将促进数据中模型的接地。我们将通过构建模型来测试和验证新工具 𝜀𝜀 与定量实验者合作，研究IgE受体(Fc RI)信号。我们将重点关注模特 𝜀𝜀 对于由有序液体组成的非均相质膜中的Fc RI-LYN相互作用 以及错乱区域和Fc RI介导的Syk激活。这些计划中的应用程序将确保我们的 软件开发活动针对有用的功能，并将提供功能演示。

PROJECT SUMMARY/ABSTRACT Rule-based modeling approaches, which are based on the principles of chemical kinetics and diffusion and enabled by an expanding armamentarium of sophisticated software tools (e.g., BioNetGen/NFsim), offer spe- cial advantages for studying the dynamics of interactions among multisite signaling proteins. Rule-based mod- els can capture the effects of polymerization-like reactions and multisite post-translational modifications over time scales of seconds to hours while incorporating constraints imposed by molecular structures. Furthermore, with a rule-based approach to model formulation, it is possible to construct and analyze larger, more compre- hensive models for cellular regulatory systems than with traditional modeling approaches because of the op- portunity to represent systems concisely and at a high level of abstraction using formal rules for biomolecular interactions. Rules can often be processed to automatically derive traditional model forms, such as a coupled system of ordinary differential equations (ODEs). However, when the system state space implied by rules is exceedingly large, the use of simulation engines based on network-free algorithms becomes necessary and model analysis is limited by the high computational cost of the stochastic simulations. In addition, in these cir- cumstances and others, parameter identification and uncertainty quantification (UQ) are extremely challenging. We will address these problems by improving the efficiency of simulation, fitting, and UQ tools and by leverag- ing distributed computing resources. Recently, we developed novel algorithms for accelerating stochastic simu- lations, a toolbox of parallelized metaheuristic optimization methods for fitting, and implementations of Markov chain Monte Carlo (MCMC) methods for Bayesian UQ. This toolbox, called PyBioNetFit (PyBNF), leverages standardized formats for defining and sharing models (e.g., core SBML and BNGL) and is compatible with var- ious simulators. Here, we propose to develop general-purpose software implementations for accelerated net- work-free (stochastic) simulation and for restructuring rule-based models (i.e., optimizing rules so as to mini- mize the number of rule-implied equations). We will also provide a new interface to CVODE and CVODES for numerical integration of ODEs, forward sensitivity analysis, and adjoint sensitivity analysis. Furthermore, we will extend the biological property specification language (BPSL) of PyBNF to make this means for formalizing qualitative data more expressive. In addition, we will add gradient-based optimization and MCMC methods to PyBNF and built-in support for Smoldyn, a simulator for (rule-based) spatial stochastic models. These im- 𝜀𝜀 provements will facilitate grounding of models in data. We will test and validate new tools by building models 𝜀𝜀 for IgE receptor (Fc RI) signaling in collaboration with quantitative experimentalists. We will focus on models 𝜀𝜀 for Fc RI-Lyn interaction within the context of a heterogeneous plasma membrane consisting of liquid ordered and disorded regions and Fc RI-mediated activation of Syk. These planned applications will ensure that our software development activities are directed at useful capabilities and will provide capability demonstrations.