面向癌症代谢系统的建模及应用研究

发布时间：2018-10-05 14:25

【摘要】：癌症是一种特殊的复杂病,是尚未攻克的难题。新陈代谢是一切生命体生存的基础,生命体的表型和它的新陈代谢紧密相关,认识癌细胞的代谢具有重要的现实意义。本文运用计算机仿真的方式对不同种类癌细胞的代谢系统进行建模,从而定性地描述癌细胞的代谢行为,为癌症研究提供有价值的线索。癌细胞的新陈代谢复杂,故而对癌细胞的代谢系统建模分析发展缓慢。由于癌细胞明显的表型特征,我们坚信其内部代谢状态必有一定规律。因而,我们借由人类全基因组网络和多维组学数据,尝试在计算机上模拟和分析癌细胞的代谢。本文完成的主要特色工作包括:(1)针对目前癌细胞代谢目标无法确定,泛癌症代谢层研究缺失等不足,给出一种在代谢层面研究泛癌症共性特征的方法,我们基于最小化反应在代谢层和基因层变化的不一致性的数学模型,定性预测癌细胞的新陈代谢显著变化情况。将该方法广泛应用到多种癌细胞上,我们成功证明或预测了泛癌症共有的代谢反应、代谢行为等特征;(2)针对特征提取算法繁多,无从选择,并且十分依赖数据特点的不足,基于癌症的代谢行为,给出一种新的特征变量提取方式,并尝试构造一棵融合代谢特征的分类树,用以识别癌症类型。将泛癌症层面给出的共有反应作为特征,分类癌症样本和正常样本获得了良好的准确率。利用癌症代谢行为的布尔关系式构造分类树,能够初步判定新样本的癌症类型。(3)针对目前敲除工作中,未考虑癌症异质性这一不足,给出一种基于特异代谢网络的,预测泛癌症致死和抑癌基因的方法。我们为每一类癌症重构出癌症特异的代谢网络,尝试利用计算机仿真基因敲除实验以预测癌细胞的两类基因。将我们的方法推广到多类癌症中,能够证明或预测某类癌症特异,或泛癌症层面具有致死、抑癌效果的两类基因,为相关湿实验提供有价值的线索。
[Abstract]:Cancer is a special complex disease that has yet to be solved. Metabolism is the basis of survival of all living organisms. The phenotype of life is closely related to its metabolism. It is of great practical significance to understand the metabolism of cancer cells. In this paper, the metabolic system of different kinds of cancer cells is modeled by computer simulation, and the metabolic behavior of cancer cells is described qualitatively, which provides valuable clues for cancer research. Cancer cell metabolism is complex, so modeling and analysis of cancer cell metabolic system is slow. Due to the obvious phenotypic characteristics of cancer cells, we firmly believe that their internal metabolic state must have a certain regularity. Therefore, we try to simulate and analyze the metabolism of cancer cells on the computer by using the human genome network and multidimensional genomics data. The main features of this paper are as follows: (1) in view of the deficiency of the current cancer cell metabolic target and the lack of research on the pan-cancer metabolic layer, a method to study the universal characteristics of pan-cancer at the metabolic level is presented. We qualitatively predict significant metabolic changes in cancer cells based on a mathematical model that minimizes the inconsistency of changes in metabolic and genetic levels. This method has been widely used in many kinds of cancer cells, and we have successfully proved or predicted the common metabolic reaction, metabolic behavior and other characteristics of pan-cancer. (2) there are many feature extraction algorithms, there is no choice, and we rely heavily on the lack of data characteristics. Based on the metabolic behavior of cancer, a new feature variable extraction method is proposed, and a classification tree is constructed to identify cancer types. Taking the common response from pan-cancer level as a feature, the classification of cancer samples and normal samples obtained a good accuracy. Using the Boolean relation of cancer metabolic behavior to construct classification tree, we can preliminarily determine the cancer type of new samples. (3) to solve the problem that cancer heterogeneity is not considered in the current knockout work, a new type of cancer classification tree based on specific metabolic network is proposed. A method for predicting pan-cancer death and tumor suppressor genes. We reconstruct cancer-specific metabolic networks for each type of cancer and try to predict two types of cancer cells using computer simulation gene knockout experiments. By extending our method to multiple types of cancer, we can prove or predict the cancer specificity of a certain type of cancer, or two kinds of genes with lethal and suppressive effect on the pan-cancer level, which provide valuable clues for the related wet experiments.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R73-3

【相似文献】