复杂信号通路模型参数估计的直接微分法

Parameter Estimation of Complicated Signal Pathways Model with Direct Derivative Method

针对细胞信号通路模型的强非线性、模型参数的不确定性以及受噪声干扰的实验数据等特点,提出一种新的参数优化算法——直接微分法,估计复杂信号转导通路模型的未知参数.该方法首先采用平滑样条拟合测量数据,然后对拟合多项式进行求导来估计模型状态变量的一阶微分,从而将动态优化问题转化为线性或非线性回归问题.以TNFα诱导NF-κB信号转导通路模型为例,利用直接微分法估计未知参数,并分析了不同因素对参数估计结果的影响.仿真结果表明:该方法能够有效估计模型的未知参数,且无需迭代求解微分方程组,降低了计算复杂度.

To overcome the strong nonlinearity of cell signal transduction pathways model,and uncertainty of model parameters,and noise caused disturbing of the experimental data,a new parameter optimization algorithm,direct derivative method（DDM）,was presented to estimate unknown parameters of complex signal transduction pathways model.In this method,the measurement data were firstly fitted with smoothing splines,and then first-order derivatives of state variables were evaluated and substituted into the model.Thus,a dynamic optimization problem was converted into a linear or nonlinear regression problem.Taking TNFα mediated NF-κB signal transduction pathways as an example,DDM was applied to estimate unknown parameters of the system model,and affections of various factors on the results of parameter estimation were also analyzed.Simulation results showed that unknown parameters could be effectively estimated using this method,and computational complexity was reduced greatly since ordinary differential equations of the system model do not need to be solved iteratively.