生物系统建模中参数估计的测量集选择--以信号转导通路模型研究为例(英文)

Measurement set selection of parameter estimation in biological system modelling—A case studyof signal transduction pathways

生物系统模型通常具有很高的维数,测量数据不完备、易受噪声污染,而且生物实验成本高,所以参数估计已经成为生物系统建模的挑战性问题之一.参数的精确估计取决于测量数据的数量和质量,因此,通过优化实验设计确定如何采集测量数据是非常重要的.针对动态系统的参数估计问题,尤其是生物反应系统,提出了一种确定富含信息的测量集选择方法,通过从设计的实验中获得测量数据,以最佳的统计质量估计系统的未知参数.该方法首先利用矩阵论的系统分析来确定估计参数所必需的测量状态的数目,再通过基于Fisher信息阵的优化实验设计决定每个测量状态的优先等级.最后,以信号转导通路模型为例,解释了该方法的优势和适用性.

Parameter estimation is a challenging problem for biological system modelling since the model is normally of high dimension, the measurement data are sparse and noisy, and the cost of experiments is high. Accurate recovery of parameters depends on the quantity and quality of measurement data. It is therefore important to know which measurements to be taken, when and how through optimal experimental design （OED）. In this paper, a method was proposed to determine the most informative measurement set for the parameter estimation of dynamic systems, in particular, biochemical reaction systems, such that the unknown parameters can be inferred with the best possible statistical quality using the data collected from the designed experiments. System analysis using matrix theory was used to examine the number of necessary measurement variables. The priority of each measurement variable was determined by optimal experimental design based on Fisher information matrix （FIM）. The applicability and advantages of the proposed method were shown through an example of a signal pathway model.