Lück Alexander, Wolf Verena
Department of Computer Science, Saarland University, Campus E 13, Saarbrücken, 66123, Germany.
BMC Syst Biol. 2016 Oct 21;10(1):98. doi: 10.1186/s12918-016-0342-8.
Discrete-state stochastic models have become a well-established approach to describe biochemical reaction networks that are influenced by the inherent randomness of cellular events. In the last years several methods for accurately approximating the statistical moments of such models have become very popular since they allow an efficient analysis of complex networks.
We propose a generalized method of moments approach for inferring the parameters of reaction networks based on a sophisticated matching of the statistical moments of the corresponding stochastic model and the sample moments of population snapshot data. The proposed parameter estimation method exploits recently developed moment-based approximations and provides estimators with desirable statistical properties when a large number of samples is available. We demonstrate the usefulness and efficiency of the inference method on two case studies.
The generalized method of moments provides accurate and fast estimations of unknown parameters of reaction networks. The accuracy increases when also moments of order higher than two are considered. In addition, the variance of the estimator decreases, when more samples are given or when higher order moments are included.
离散状态随机模型已成为一种成熟的方法，用于描述受细胞事件固有随机性影响的生化反应网络。近年来，几种准确逼近此类模型统计矩的方法变得非常流行，因为它们能够对复杂网络进行高效分析。
我们提出了一种广义矩方法，用于基于相应随机模型的统计矩与群体快照数据的样本矩的精确匹配来推断反应网络的参数。所提出的参数估计方法利用了最近开发的基于矩的近似方法，并在有大量样本时提供具有理想统计特性的估计量。我们通过两个案例研究证明了该推断方法的实用性和效率。
广义矩方法能够对反应网络的未知参数进行准确且快速的估计。当考虑高于二阶的矩时，准确性会提高。此外，当提供更多样本或包含更高阶矩时，估计量的方差会减小。
