Cavelier German, Anastassiou Dimitris
Genomic Information Systems Laboratory, Department of Electrical Engineering, Columbia University, New York, New York 10027, USA.
Proteins. 2004 May 1;55(2):339-50. doi: 10.1002/prot.20056.
Finding the causality and strength of connectivity in transcriptional regulatory networks from time-series data will provide a powerful tool for the analysis of cellular states. Presented here is the design of tools for the evaluation of the network's model structure and parameters. The most effective tools are found to be based on evolution strategies. We evaluate models of increasing complexity, from lumped, algebraic phenomenological models to Hill functions and thermodynamically derived functions. These last functions provide the free energies of binding of transcription factors to their operators, as well as cooperativity energies. Optimization results based on published experimental data from a synthetic network in Escherichia coli are presented. The free energies of binding and cooperativity found by our tools are in the same physiological ranges as those experimentally derived in the bacteriophage lambda system. We also use time-series data from high-density oligonucleotide microarrays of yeast meiotic expression patterns. The algorithm appropriately finds the parameters of pairs of regulated regulatory yeast genes, showing that for related genes an overall reasonable computation effort is sufficient to find the strength and causality of the connectivity of large numbers of them.
从时间序列数据中找出转录调控网络中的因果关系和连接强度，将为细胞状态分析提供一个强大的工具。本文介绍了用于评估网络模型结构和参数的工具设计。发现最有效的工具是基于进化策略的。我们评估了从集总代数现象学模型到希尔函数和热力学推导函数等复杂度不断增加的模型。这些最后的函数提供了转录因子与其操纵子结合的自由能以及协同能。展示了基于已发表的大肠杆菌合成网络实验数据的优化结果。我们的工具所发现的结合自由能和协同能与在噬菌体λ系统中通过实验得出的处于相同的生理范围内。我们还使用了来自酵母减数分裂表达模式的高密度寡核苷酸微阵列的时间序列数据。该算法适当地找到了成对调控的酵母基因的参数，表明对于相关基因，总体上合理的计算量足以找出大量基因连接的强度和因果关系。
