基于蛋白质互作网络的疾病相关miRNA挖掘方法的研究

发布时间：2018-03-20 20:45

本文选题：miRNA　切入点：疾病　出处：《哈尔滨工业大学》2008年硕士论文　论文类型：学位论文

【摘要】：在人类胚胎发育和疾病发生等过程中,miRNA扮演着重要的调控角色。而随着miRNA研究的深入,有关miRNA的生物学数据正迅速增多。由此,通过寻找生物学数据之间的联系,生物信息学研究者开始考虑从纯数据角度来研究miRNA与疾病的关系,以填补传统实验方法研究的缺陷。疾病相关miRNA挖掘的研究将对疾病的诊断治疗、药物靶标的寻找、药物的筛选等过程提供重要的指导作用。本文主要研究基于蛋白质互作网络的疾病相关miRNA的挖掘方法。首先,分析了miRNA、疾病相关的生物学数据的特征,确定了一套有效的生物学数据作为本文研究的基础。然后,利用选择好的生物学数据,定义并构建了疾病相关的蛋白质互作网络,并对该网络作了三类拓扑参数(度,簇系数,最短路径条数)的统计分析。接着,利用miRNA的靶点数据特征,建立了miRNA与疾病相关的蛋白质互作网络之间的联系。最后,把贝叶斯后验概率的数学模型应用于疾病相关的蛋白质互作网络的拓扑参数分析上,成功实现了预测疾病相关miRNA的目的。此外,本文在算法设计的基础上,完成了乳腺癌相关miRNA的预测工作,并对算法的有效性做了多次四倍交叉验证。结果显示,在正确指数为0.32的情况下,算法预测准确率能够达到70%左右,从而有效地论证了本文基于蛋白质互作网络的疾病相关miRNA挖掘算法的正确性。
[Abstract]:MiRNAs play an important role in the regulation of human embryonic development and disease. With the development of miRNA, the biological data about miRNA are increasing rapidly. Bioinformatics researchers begin to consider studying the relationship between miRNA and disease from the perspective of pure data to fill the defects of traditional experimental methods. The research of disease-related miRNA mining will be used to diagnose and treat diseases and find drug targets. Drug screening and other processes provide important guidance. In this paper, we mainly study the method of miRNA mining based on protein interaction network. Firstly, we analyze the characteristics of miRNAs and disease-related biological data, and determine a set of effective biological data as the basis of this study. Using the selected biological data, the disease related protein interaction network is defined and constructed, and three kinds of topological parameters (degree, cluster coefficient, shortest path number) are analyzed. Based on the target data of miRNA, the relationship between miRNA and disease-related protein interaction networks is established. Finally, the mathematical model of Bayesian posteriori probability is applied to the topological parameter analysis of disease-related protein interaction networks. The goal of predicting disease related miRNA has been successfully achieved. In addition, on the basis of the algorithm design, the prediction work of breast cancer related miRNA is completed, and the validity of the algorithm is verified four times. The results show that the correct index is 0.32. The prediction accuracy of the algorithm can reach about 70%, which effectively proves the correctness of the disease related miRNA mining algorithm based on protein interaction network in this paper.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2008
【分类号】：R341

【参考文献】