MicroRNA功能的生物信息学分析及相关平台的建立

【作者】 颜兴起；

【导师】 袁建刚； 李亦学； 强伯勤； 彭小忠；

【作者基本信息】 中国协和医科大学 ， 生物化学与分子生物学， 2008， 博士

【摘要】 miRNA(microRNA)是一类重要的非编码RNA,通过在转录后水平上抑制靶基因的mRNA翻译或降解靶基因mRNA来发挥作用。根据miRBase数据库,到目前为止,已经发现了678个人类的microRNA基因,并且有证据表明在人类中尚有大量的microRNA有待发现,microRNA的总数可能会更多。目前认为microRNA参与很多重要的生物学过程,如转录因子调控网络,发育过程中的时序控制,神经突触形成,细胞增殖,细胞死亡,细胞分化等。micmRNA的研究成为生命科学中的研究热点,本文就利用生物信息方法研究microRNA的功能进行了初步探索。目前有多种计算生物学的工具用来预测哺动物中micmRNA的靶基因(miRanda,TargetScan,PicTar),但得到实验验证的靶基因数目远远小于通过计算方法预测的数目。现实的情况是,尽管发现了大量的microRNA,然而寻找microRNA靶基因的步伐相对迟缓。发现microRNA的靶基因将极大地促进microRNA的功能研究,因此本文第一部分提出了一个用于改进microRNA靶基因预测效率的整合机器学习算法,算法在训练集上、FMRP相关mRNA数据集上都获得了较好的效果。进一步选择miR-9作为测试对象,随机抽取16个预测靶基因进行荧光素酶报告基因验证实验,结果显示,有10个预测靶基因的结果与预期相符。通过以上结果,我们可以得出结论,我们开发的整合算法可以有效地提高现有microRNA靶基因预测算法的准确率。近年来,基因芯片技术被用来研究microRNA基因的调节功能。然而如何将基因的表达谱和microRNA的表达谱合并在一起进行分析,依然没有被很好地解决。本文第二部分提出了一个新的概念,即in-silico MRPs(in-silico MicroRNA RegulatoryProfiles),将两者结合在一起,首次用来从基因组水平上描述microRNA的调节功能。我们首先为三个物种,即人(157 microRNAs×13041 mRNAs),大鼠(152microRNAs×5108 mRNAs)与小鼠(72 microRNAs×10729),分别构建了各自的in-silicoMRPs矩阵。通过验证,构建的in-silico MRP确实能够真实、有效地反映一个microRNA基因的调控作用。我们进一步以in-silico MRP为基础,从基因组水平上获得了两组新的数据:在人中有约36%的microRNA倾向于降解其靶基因,及三个物种间有大约42%的microRNA其调控靶基因的能力保守。为了方便其他的研究者使用in-silico MRP,我们建立了一个网站,提供已经计算好的in-silico MRP下载,并提供在线计算in-silico MRP的能力,为自己有数据而又不方便公布的研究者提供便利。目前,microRNA相关功能分析软件很多,然而这些软件和资源散布在各处,联系松散,不方便研究者使用。本文的第三部分描述了一个专门为microRNA功能分析设计的R语言程序包—miRE。miRE具有良好的图形化界面,可以辅助研究者进行microRNA相关功能分析,如浏览microRNA的生物学注释,预测靶基因,文献挖掘及microRNA基因邻近特征分析等。综上所述,本文提供了一种改良的整合机器算法来提高预测靶基因的准确率,提供一种整合miRNA/mRNA表达谱的方法,并开发了一种平台性软件miRE,为microRNA相关的功能分析提供了有力的工具。更多还原

【Abstract】 MicroRNAs are a class of small endogenous non-coding RNAs which play important regulatory roles mainly by posttranscriptional depression. To date, there are 678 human microRNA sequences, according to the miRBase sequence database (Release 11, released April, 2008), have been identified. Moreover, there exists a hypothesis that the total number of human microRNAs will be much larger. MicroRNAs have been thought to be involved in many biological processes, such as transcriptional gene regulatory network, developmental timing, neuronal synapses formation, cell proliferation, cell death, and differentiation.Although there are several computational programs served to predict miRNA targets in mammals (miRanda, TargetScan, PicTar), the fact is that there are many predictions while only few of them have been biologically validated. Finding miRNA target genes will help a lot to understand their biological functions. Unfortunately, the prediction of miRNA target genes is more challenging. In the first section of the dissertation, we developed an ensemble machine learning algorithm which helps to improve the prediction of miRNA targets. The performance was evaluated in the training set and in FMRP associated mRNAs. Moreover, using human mir-9 as a test case, our classification was validated in 10 of 16 transcripts tested. From the results we got, we could make the conclusion that our ensemble algorithm can improve the prediction of microRNA targets efficiently.Microarray-based studies have been used to investigate the regulatory properties of microRNAs. However, to integrate both microRNA and mRNA expression profiles to study sophisticated microRNA functions has not been well addressed. In the second section, we proposed a novel concept of in-silico MicroRNA Regulatory Profiles (in-silico MRPs) for quantifying the microRNA’s regulatory properties. The in-silico MRPs were constructed for human (involving 157 microRNAs×13041 mRNAs), mouse (72 microRNAs×10729 mRNAs), and rat (152 microRNAs×5108 mRNAs). We proved that the in-silico MRPs are reliable by comparing our in-silico MRPs with the real gene expression profiles after over expressing or knocking down a specific microRNA. Given these in-silico MRPs, we have explored microRNAs that are likely to function through degrading their targets, the results showed about 36% of human microRNA fall into this category. 42% of shared miroRNAs across the three species were found to retain their regulatory ability through the evolution. A web server (http://www.biosino.org/VirtualOverExpressWebApp/) has been provided where our constructed in-silico MRPs can be downloaded for research use, and customerized in-silico MRPs can be generated if users load their real experimental data. A few other possible applications of the web service were suggested.Numerous computational tools have been developed for studying the function of this class of small nucleotides, and resources and tools are quite useful for studying the function of microRNA; however, they are distributed in different websites and loosely connected. So far, there is no tool to provide an open-source, user-friendly, integrated interface for microRNA related analysis. We have developed an R package, named miRE, which provides a graphical user interface for conducting microRNA related analysis.更多还原