COMPUTATIONAL TOOLS FOR RULE-BASED MODELING OF BIOCHEMICAL SYSTEMS

描述(申请人提供)：生化系统的一个共同特征，特别是那些蛋白质-蛋白质相互作用突出的系统，是组合的复杂性，当相对较少的相互作用有可能产生大量不同的化学物种和反应时就会出现这种情况。对于一个以组合复杂性为标志的系统，如果数学模型要全面考虑生物分子位置水平上的相互作用的结果，那么手动指定数学模型的每一项的传统方法是不可能的。这项提议的主要目标是能够快速开发信号转导系统的机械模型，以逻辑一致的方式尽可能全面地解释蛋白质-蛋白质相互作用的结果。为了实现这一目标，我们将开发模型规范和检查工具。这些工具将基于涉及使用图形来表示蛋白质以及使用图形重写规则来表示蛋白质-蛋白质相互作用的方法。规则是可视的，很像图形化的交互地图。每条规则规定了一种由生物分子相互作用引起的结合/酶反应类型，并确定了反应物的特征。规则可以通过图形重写的过程自动解释，以获得各种类型的数学模型。因此，规则使生物分子相互作用的精确和可理解的可视化成为可能。重要的是，一组规则是组合的，因为每个规则都可以单独指定和细化。另一方面，传统模型中的方程通常是相互关联的，改变关于蛋白质-蛋白质相互作用的假设可能需要对多个方程进行大量修改。拟议工作的第二部分旨在演示基于规则的建模的实用性。为了确保和证明我们的工具是有用的，我们将为一些生物系统开发模型。我们还将演示如何将我们的工具与数据库资源一起用作高吞吐量建模管道的一部分。我们希望实现的一个重要能力是能够对已知的人类信号转导系统的很大一部分进行建模。

DESCRIPTION (provided by applicant): A common feature of biochemical systems, especially those in which protein-protein interactions are prominent, is combinatorial complexity, which is present whenever a relatively small number of interactions have the potential to generate a much larger number of distinct chemical species and reactions. For a system marked by combinatorial complexity, the conventional approach of manually specifying each term of a mathematical model is impossible if the model is to account comprehensively for the consequences of interactions at the level of biomolecular sites. The primary goal of this proposal is to enable rapid development of mechanistic models of signal-transduction systems that account as completely as possible for the consequences of protein-protein interactions in a logically consistent way. To achieve this goal, we will develop tools for model specification and checking. These tools will be based on methods that involve the use of graphs to represent proteins and graph rewriting rules to represent protein-protein interactions. The rules are visual, much like diagrammatic interaction maps. Each rule specifies a type of binding/enzymatic reaction that arises from a biomolecular interaction and identifies features of reactants. Rules can be interpreted automatically, through procedures of graph rewriting, to obtain various types of mathematical models. Thus, rules enable precise and comprehensible visualization of biomolecular interactions. Importantly, a set of rules is compositional, in that each rule may be specified and refined independently. Equations in a conventional model on the other hand are typically interrelated, and changing an assumption about a protein-protein interaction may require numerous modifications of multiple equations. The second part of the proposed work is aimed at demonstrating the practicality of rule-based modeling. To ensure and demonstrate that our tools are useful, we will develop models for a number of biological systems. We will also demonstrate how our tools, together with database resources, can be used as part of a high- throughput modeling pipeline. An important capability we wish to achieve is the ability to model a significant fraction of the known human signal-transduction systems.