细胞信号转导网络的全局灵敏度分析

Global Sensitivity Analysis of Cell Signal Transduction Network

灵敏度分析是定性定量研究系统生物学模型不确定性的主要方法,可以有效确定主导参数,减小模型不确定性,提高优化效率。然而如何快速有效地评估参数灵敏度现已成为系统生物学模型优化的难题。针对传统的全局灵敏度方法对多参数复杂系统生物学模型分析的不足,采用Kriging代理模型法,结合改进的基于方差的Sobol法,建立采用代理模型法的Sobol定量全局灵敏度分析法,实现复杂模型参数灵敏度分析的快速定量化评估。选用基于Smad蛋白的TGF-B信号转导网络进行实例研究,以细胞核内磷酸化Smad复合物为目标函数评价模拟效果。研究表明,上述方法在实现定量全局灵敏度分析的同时降低了模型运行时耗,提高模型评估效率,且与传统Sobol法具有同样的评估效果。

Sensitivity analysis is the main method of qualitative and quantitative study of model uncertainty of biological systems, which can effectively determine the dominant parameters, reduce the model calibration uncertainty and enhance the model optimization efficiency. However, how to quickly and effectively validate a model and identify the dominant parameters for biological system model is a problem to achieve the parameter optimization. There are some shortcomings for the classical method of global multi-parameter analysis of complex biological model systems. And using Kriging meta-modeling approach combined with the improved method which is based on the variance of Sobol can achieve the rapid and quantitative evaluation for complex model parameter sensitivity analysis. Taking TGF -βsignal transduction network based on Smad as a case study, it can select phosphorylated Smad complex in the cytoplasm to assess the model as the output response for sensitivity analysis. The results show that the new method can not only achieve the quantification of the sensitivity, but also reduce the computational cost.