【作者】 张志勇；

【导师】 陈洛南；

【作者基本信息】 上海大学 ， 应用数学， 2013， 博士

【摘要】 以生物学为代表的生命科学和以复杂系统和复杂性研究为核心的复杂性科学都是二十一世纪最重要的研究领域，系统生物学则处于这两个学科的重叠区域。一般来说，系统生物学尝试在系统水平上理解生物对象的功能运作，这需要综合使用不同种类的方法以实现不同来源数据的整合，最终达成对生物系统涌现特性的刻画和建模。在本论文中，我们选择了两个课题进行研究，一个是时序列基因表达数据上的双聚类算法，另一个是哺乳类动物SCN昼夜节律的达尔文式模型。这两个课题的研究有着不同的学术路径，我们希望由此展示我们对系统生物学研究路径的一些思考。第一个课题的背景是基因调控网的推断，它展示了我们对计算系统生物学研究途径的认识。时序列基因表达数据的双聚类是要寻找出在特定时间段内具有表达相关性的基因族。我们受到基于后缀树的双聚类算法启发，首先把基因表达数据表示为加权字符串，然后根据这些加权字符串生成广义加权后缀树，我们发现，这个广义加权后缀树中的内部节点给出了双聚类结果，而这一算法的时间复杂度和空间复杂度都是线性的。另外，我们还将算法推广到更复杂的应用背景下，包括容许错配、容许引入比对得分矩阵等情形，都有很简洁的方案。这一算法较好地解决了时序列基因表达数据的双聚类问题，为后续的数据处理工作提供了良好的基础。第二个课题的背景是复杂系统的功能机制，我们猜测达尔文式的机制和过程可能在复杂系统的功能运作中具有相当的普遍性。而我们以哺乳动物SCN昼夜节律为例，具体展示了这一机制和过程如何在生物钟运作中是可能的。昼夜节律现象可能是最普遍的生物现象之一，出现在从生化过程到细胞、到组织、到个体、到种群、到生态群落的不同生物层次上。我们引入经济学中的机制设计框架，论证SCN在运转中可以采用类似于自然选择机制下进化的达尔文式过程来实现SCN的授时功能，特别地，我们在这一框架下给出光线诱导下的相位响应曲线的全新解释。更多还原

【Abstract】 Life Science and Science of Complexity are both the most important researchareas in the21st century. Systems biology is the intersection of these two areas. Gen-erally speaking, Systems Biology aims at understanding the function of biologicalobjects at systems level, which needs to synthesize different methods to fulfill the in-tegration of data from different sources and finally models emergency phenomenon inbiological system.In the thesis, we will study two problems. One is the biclustering algorithmfor time-series gene expression data, the other is the Darwinian modeling circadianrhythm of the mammalian SCN. We will study these two problems with different ap-proaches, aiming at showing our thinking on the study approaches to problems inSystems Biology.The research background of the first problem is gene regulatory network infer-ence. Biclustering algorithms of time-series gene expression data try to find out thosegene groups within each group the expression of the genes are correlated within a spe-cific time interval. Inspired by a biclustering algorithm based on suffix tree, we pro-pose a new one. The gene expression data are firstly represented as weighted strings,then transformed into general weighted suffix tree. We find out the inner nodes ofthe given general weighted suffix tree give the bicluster results. Both time complex-ity and space complexity are linear. Furthermore, the algorithm can be generalizedsmoothly to more complex application backgrounds, including mismatch, alignmentscore matrix, etc. The algorithm provides a good solution to the gene expression databiclustering problem and lays a sound foundation for further research.The research background of the second problem is the mechanism of functionof complex systems. We conjecture that the Darwinian mechanism and Darwinianprocess are common in complex systems. We take circadian rhythm of mammalian SCN as an example to illustrate how this mechanism and process could play the rolein SCN. Circadian rhythm is perhaps one of the most common biological phenomenonemerging on different biological levels, ranging from biomedical processes to cell,organism, individual, population and ecological system. We introduce the frameworkof mechanism design problem rooted in economics into our model, and demonstrateSCN cells can fulfill their time-telling function by adopting the Darwinian processwhich is similar to the process of Evolution by Natural Selection. In particular, wegive a brand-new explanation of the light-induced Phase Response Curve under thismodel.更多还原