基于机器学习的microRNA基因预测

发布时间：2018-04-03 01:31

  本文选题：microRNA　切入点：pre-microRNA　出处：《河北工业大学》2011年硕士论文

【摘要】：microRNA是一种单链的非编码小分子RNA,长约20-24个核苷酸,它由长度约为70个核苷酸的microRNA前体(pre-microRNA)经过具有RNaseIII活性的Dicer和Dicer-like-1内切核酸酶加工形成,microRNA通过诱导靶mRNA剪切或者抑制其翻译来调控基因表达的功能。人类有近三分之一的基因都受到microRNA的调控,它对生物的细胞增殖分化、细胞死亡、早期发育、代谢活动等生物过程有着重要的调控作用,研究表明,它与癌症也有着紧密的联系,对microRNA的研究有助于人们了解基因间的网络调控关系,更有助于对基因功能的研究以及生物的进化探索。虽然microRNA广泛存在于55个物种之中,目前被鉴定出来的microRNA数量比实际存在的要少的多,还有大量的microRNA有待发现。因此对microRNA的进行预测具有重要的意义。 目前已知的microRNA预测主要有两种方式,cDNA克隆预测和计算预测。前者是microRNA早期主要的预测方式,这种方式直接、可靠,但是很难克隆出在不同时期表达,或者只在特定的组织或细胞系中表达的microRNA。计算预测则不会受到microRNA在表达时间、表达水平或组织特异性的影响,从而可以弥补cDNA克隆测序的不足。 本文基于机器学习提出了一种称为ACO+SVM的microRNA预测方法,由于pre-microRNA的序列较长,并且可以折叠形成stem-loop结构,将pre-microRNA序列和结构特征结合起来提取相应属性信息。本文通过已知的阳性和阴性pre-microRNA构建区分二者的分类器,由于支持向量机(Support Vector Machines,SVM)在逼近和泛化能力方面具有良好的特性,因此本文microRNA的预测方法采用SVM训练分类器,考虑到SVM分类器的性能受核函数和相关参数的影响很大,采用蚁群算法(Ant Colony Optimization,ACO)搜索SVM的相关参数,以构建无偏、且同时具有较高敏感性和特异性的分类器。实验结果表明该方法不仅可以有效的鉴别人类真假pre-microRNA,而且在其他多个物种的预测上具有较高的准确度,与其他同类方法相比具有更好的敏感性和特异性。
[Abstract]:MicroRNA is a single-stranded, non-coding small molecule with about 20-24 nucleotides.It is processed from 70 nucleotides of microRNA precursor pre-microRNAs through Dicer and Dicer-like-1 endonucleases with RNaseIII activity to form the ability of microRNAs to regulate gene expression by inducing target mRNA splicing or inhibiting its translation.Nearly 1/3 of human genes are regulated by microRNA, which plays an important role in biological processes such as cell proliferation and differentiation, cell death, early development, metabolic activity, etc.It is also closely related to cancer. The study of microRNA is helpful to understand the network regulation between genes, to study the function of genes and to explore the evolution of organisms.Although microRNA is widespread in 55 species, the number of microRNA identified is much smaller than the actual number, and a large number of microRNA remains to be discovered.Therefore, the prediction of microRNA is of great significance.At present, there are two main methods of microRNA cloning prediction and computational prediction.The former is the main early prediction method of microRNA, which is direct and reliable, but it is difficult to clone microRNAs expressed at different stages or only in specific tissues or cell lines.Calculation and prediction will not be affected by the expression time, expression level or tissue specificity of microRNA, which can make up for the deficiency of cDNA cloning and sequencing.In this paper, a microRNA prediction method called ACO SVM is proposed based on machine learning. Because the pre-microRNA sequence is long and can be folded into stem-loop structure, the pre-microRNA sequence and structure feature are combined to extract the corresponding attribute information.In this paper, a classifier is constructed by using known positive and negative pre-microRNA. Because the support vector machine (SVM) has good properties in approximation and generalization, SVM is used to train the classifier in this microRNA prediction method.Considering that the performance of SVM classifier is greatly affected by kernel function and related parameters, ant colony algorithm (Ant Colony optimization) is used to search the relevant parameters of SVM to construct an unbiased classifier with high sensitivity and specificity.The experimental results show that this method not only can effectively identify human pre-microRNAs, but also has higher accuracy in predicting other species, and has better sensitivity and specificity than other similar methods.
【学位授予单位】：河北工业大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TP181;R346

【参考文献】

相关期刊论文 前4条

1 杨良怀;吕丕明;陈立军;邓明华;;k-gram方法识别microRNA前体[J];生物化学与生物物理进展;2007年02期

2 张春霆;生物信息学的现状与展望[J];世界科技研究与发展;2000年06期

3 徐宁,李春光,张健,虞厥邦;几种现代优化算法的比较研究[J];系统工程与电子技术;2002年12期

4 侯妍妍;应晓敏;李伍举;;microRNA计算发现方法的研究进展[J];遗传;2008年06期

相关博士学位论文 前1条

1 秦玉平;基于支持向量机的文本分类算法研究[D];大连理工大学;2008年

，

本文编号：1703040