采用Levenberg-Marquardt算法的信号转导通路模型参数估计

Estimating the Parameters of Signal Transduction Pathways with Levenberg-Marquardt Algorithm

建立信号转导通路的数学模型,以便定量对其进行研究是系统生物学的一个基本任务,但是,参与信号转导通路的蛋白质众多,相互关系复杂,以及信号转导通路表现出的强非线性特性使模型未知参数的估计十分困难。以肿瘤坏死因子α(Tumor necrosisα,TNFα)诱导的细胞核因子κB(Nuclear Factor-κB,NF-κB)信号转导通路为例,研究了基于Levenberg-Marquardt算法在信号转导通路模型参数估计中的应用,该方法包括两部分,首先采用正交化方法确定未知参数的可辨识性,然后采用Levenberg-Marquardt优化算法在给定的测量时间序列下,估计参数值。仿真结果证明该方法能较好地估计信号转导通路模型的未知参数并且具有一定的稳定性。

The modeling of signal transduction pathways is a task of systems biology. However, such a task is very difficult because of the structure complexity, the strong nonlinearity of signaling pathways and the noised and incomplete measurements. The Levenberg-Marquardt algorithm（LM algorithm） is applied to estimate the unknown parameters of the signaling pathways. With this method, the identifiahility of unknown parameters is appraised, and the sensitivity equations of original model are evaluated. Then we append the sensitivity equations to the original model in order to form the augmented model, and we apply the Levenberg-Marquardt algorithm to the augmented model in order to estimate parameters. TNFα mediated NF-kB signaling pathway is taken as an example to illustrate the effectiveness of this method, and the simulation results are given.