Sriyudthsak Kansuporn, Sawada Yuji, Chiba Yukako, Yamashita Yui, Kanaya Shigehiko, Onouchi Hitoshi, Fujiwara Toru, Naito Satoshi, Voit Ebernard O, Shiraishi Fumihide, Hirai Masami Yokota
BMC Syst Biol. 2014;8 Suppl 5(Suppl 5):S4. doi: 10.1186/1752-0509-8-S5-S4. Epub 2014 Dec 12.
Progress in systems biology offers sophisticated approaches toward a comprehensive understanding of biological systems. Yet, computational analyses are held back due to difficulties in determining suitable model parameter values from experimental data which naturally are subject to biological fluctuations. The data may also be corrupted by experimental uncertainties and sometimes do not contain all information regarding variables that cannot be measured for technical reasons.
We show here a streamlined approach for the construction of a coarse model that allows us to set up dynamic models with minimal input information. The approach uses a hybrid between a pure mass action system and a generalized mass action (GMA) system in the framework of biochemical systems theory (BST) with rate constants of 1, normal kinetic orders of 1, and -0.5 and 0.5 for inhibitory and activating effects, named Unity (U)-system. The U-system model does not necessarily fit all data well but is often sufficient for predicting metabolic behavior of metabolites which cannot be simultaneously measured, identifying inconsistencies between experimental data and the assumed underlying pathway structure, as well as predicting system responses to a modification of gene or enzyme. The U-system approach was validated with small, generic systems and implemented to model a large-scale metabolic reaction network of a higher plant, Arabidopsis. The dynamic behaviors obtained by predictive simulations agreed with actually available metabolomic time-series data, identified probable errors in the experimental datasets, and estimated probable behavior of unmeasurable metabolites in a qualitative manner. The model could also predict metabolic responses of Arabidopsis with altered network structures due to genetic modification.
The U-system approach can effectively predict metabolic behaviors and responses based on structures of an alleged metabolic reaction network. Thus, it can be a useful first-line tool of data analysis, model diagnostics and aid the design of next-step experiments.
系统生物学的进展为全面理解生物系统提供了复杂的方法。然而，由于难以从自然存在生物波动的实验数据中确定合适的模型参数值，计算分析受到了阻碍。数据还可能因实验不确定性而被破坏，有时由于技术原因不包含关于无法测量变量的所有信息。
我们在此展示了一种构建粗粒度模型的简化方法，该方法使我们能够以最少的输入信息建立动态模型。该方法在生化系统理论（BST）框架下使用了纯质量作用系统和广义质量作用（GMA）系统的混合形式，速率常数为1，正常动力学级数为1，抑制和激活作用分别为 -0.5和0.5，称为统一（U）系统。U系统模型不一定能很好地拟合所有数据，但通常足以预测无法同时测量的代谢物的代谢行为，识别实验数据与假定的潜在途径结构之间的不一致，以及预测系统对基因或酶修饰的反应。U系统方法在小型通用系统中得到了验证，并应用于对高等植物拟南芥的大规模代谢反应网络进行建模。预测模拟得到的动态行为与实际可用的代谢组学时间序列数据一致，识别了实验数据集中可能存在的误差，并定性估计了不可测量代谢物的可能行为。该模型还可以预测由于基因改造而导致网络结构改变的拟南芥的代谢反应。
U系统方法可以根据所谓的代谢反应网络结构有效地预测代谢行为和反应。因此，它可以成为数据分析、模型诊断的有用一线工具，并有助于设计下一步实验。
