转录因子和miRNA在复杂疾病中的共调控基因网络研究

发布时间：2018-05-03 13:24

本文选题：复杂疾病 + 转录因子　；参考：《西安理工大学》2017年硕士论文

【摘要】：复杂疾病的发生受到多个基因的调控,一直是生物医学研究的重点和难点。现代生物学和实验技术的不断发展,为深入研究基因调控机制创造了条件。研究复杂疾病的基因调控网络,对于揭示复杂疾病内部复杂的生命现象和调控规律,诊断和治疗复杂疾病具有较大的推动作用。本文遵循系统生物学和分子生物学的思想,采用生物信息学的方法,研究转录因子(Transcription Factor,TF)和microRNA (miRNA)参与调控的基因网络的构建以及基因网络的动力学机制。首先,介绍了构建复杂疾病相关的转录因子和microRNA共调控基因网络的生物信息学方法、原理以及相关数据库。然后,讨论了两种转录过程的动力学模型,提出一种转录因子和microRNA共调控的前馈环动力学模型。最后,将各种方法、数据库和前馈环动力学模型应用于胰腺癌数据。本课题利用倍数分析和精确检验对患病和正常两种样本数据进行差异表达分析。通过加权基因关联网络分析获得差异基因和差异microRNA的共表达,来预测microRNA对基因的调控关系。利用差异基因和位置权重模型匹配,预测调控基因的转录因子。对TransmiR和ENCODE数据库中调控差异microRNA的转录因子关系取并集。整合得到的调控关系,构建转录因子和microRNA共调控的基因网络,获得134个前馈环模体。使用微分方程组对前馈环的转录机制进行建模和定量分析。采用高斯过程描述隐转录因子的表达活性,基于贝叶斯框架对前馈环动力学模型进行推导,分别采用单目标文化遗传算法和同时考虑基因表达值及其梯度的多目标文化遗传算法,对动力学模型的参数和核函数的超参数进行迭代优化求解。仿真实验结果表明,本课题提出的方法可以较好地估计模型参数以及隐转录因子的活性。改进的多目标优化算法相较于单目标优化算法鲁棒性更强,降低了模型参数的估计误差,提高了隐转录因子的估计精度。
[Abstract]:The occurrence of complex diseases is regulated by multiple genes, which has been the focus and difficulty of biomedical research. The continuous development of modern biology and experimental technology has created conditions for further study of gene regulation mechanism. The study of gene regulation network of complex diseases is helpful to reveal the complex life phenomena and regulation rules within complex diseases and to diagnose and treat complex diseases. In accordance with the ideas of systems biology and molecular biology, this paper studies the construction of gene network and the dynamic mechanism of gene network regulated by transcription factor (TFF) and microRNA miRNAs by means of bioinformatics. Firstly, the bioinformatics methods, principles and related databases for the construction of complex disease-related transcription factors and microRNA coregulatory gene networks are introduced. Then, the kinetic models of two kinds of transcription processes are discussed, and a feedforward loop kinetic model of co-regulation of transcription factors and microRNA is proposed. Finally, various methods, databases and feedforward loop dynamics models are applied to pancreatic cancer data. In this paper, we use multiple analysis and accurate test to analyze the differential expression of two kinds of sample data. The co-expression of differentially expressed genes and differential microRNA was obtained by weighted gene association network analysis to predict the regulatory relationship of microRNA to genes. The transcriptional factors of regulatory genes were predicted by matching differential gene and position weight model. The transcriptional factor relationships in TransmiR and ENCODE databases regulating differential microRNA were merged. The gene network of transcription factor and microRNA was constructed, and 134 feedforward ring motifs were obtained. The transcription mechanism of feedforward loop is modeled and quantitatively analyzed by differential equations. Gao Si process is used to describe the expression activity of hidden transcription factors, and the dynamic model of feedforward loop is derived based on Bayesian framework. The single-objective cultural genetic algorithm and the multi-objective cultural genetic algorithm considering the gene expression value and its gradient are used to optimize the parameters of the dynamic model and the super-parameters of the kernel function. The simulation results show that the proposed method can estimate the model parameters and the activity of hidden transcription factors. The improved multi-objective optimization algorithm is more robust than the single-objective optimization algorithm, which reduces the estimation error of the model parameters and improves the estimation accuracy of the hidden transcription factors.
【学位授予单位】：西安理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R3416

【参考文献】