Ainscow E K, Brand M D
Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, CB2 1QW, U.K.
J Theor Biol. 1998 Sep 21;194(2):223-33. doi: 10.1006/jtbi.1998.0760.
The errors associated with experimental application of metabolic control analysis are difficult to assess. In this paper, we give examples where Monte-Carlo simulations of published experimental data are used in error analysis. Data was simulated according to the mean and error obtained from experimental measurements and the simulated data was used to calculate control coefficients. Repeating the simulation 500 times allowed an estimate to be made of the error implicit in the calculated control coefficients. In the first example, state 4 respiration of isolated mitochondria, Monte-Carlo simulations based on the system elasticities were performed. The simulations gave error estimates similar to the values reported within the original paper and those derived from a sensitivity analysis of the elasticities. This demonstrated the validity of the method. In the second example, state 3 respiration of isolated mitochondria, Monte-Carlo simulations were based on measurements of intermediates and fluxes. A key feature of this simulation was that the distribution of the simulated control coefficients did not follow a normal distribution, despite simulation of the original data being based on normal distributions. Consequently, the error calculated using simulation was greater and more realistic than the error calculated directly by averaging the original results. The Monte-Carlo simulations are also demonstrated to be useful in experimental design. The individual data points that should be repeated in order to reduce the error in the control coefficients can be highlighted.
代谢控制分析实验应用中的误差难以评估。在本文中，我们给出了一些例子，其中已发表实验数据的蒙特卡罗模拟被用于误差分析。根据从实验测量中获得的均值和误差来模拟数据，并使用模拟数据计算控制系数。重复模拟500次能够对计算出的控制系数中隐含的误差进行估计。在第一个例子中，对分离线粒体的状态4呼吸作用进行了基于系统弹性的蒙特卡罗模拟。模拟得出的误差估计与原始论文中报告的值以及从弹性敏感性分析得出的值相似。这证明了该方法的有效性。在第二个例子中，对分离线粒体的状态3呼吸作用进行了基于中间体和通量测量的蒙特卡罗模拟。该模拟的一个关键特征是，尽管原始数据的模拟基于正态分布，但模拟控制系数的分布并不遵循正态分布。因此，使用模拟计算出的误差比直接对原始结果求平均值计算出的误差更大且更符合实际情况。蒙特卡罗模拟在实验设计中也被证明是有用的。为了减少控制系数中的误差而应重复的各个数据点可以被突出显示。
