基于小波变换技术预测DNA序列编码区

【作者】 王玉；

【导师】 饶妮妮；

【作者基本信息】 电子科技大学 ， 生物医学工程， 2006， 硕士

【摘要】 随着人类基因组计划的发展,近年来GenBank里的碱基数目呈指数增长,如何从大量的数据中挖掘出有用的生物信息是生物信息学领域今后几十年都需要致力解决的问题,其中基因识别问题——通过计算的方法识别DNA序列中编码蛋白质的基因更是迫切需要解决的研究课题之一。目前基因识别常用的方法有:人工神经网络方法、复杂度分析方法、傅立叶分析方法、密码学方法和统计学方法等。概括起来,基因预测方法大致分为两类。一类是基于编码区所具有的独特信号,比如起始密码子、终止密码子等;另一类是基于编码区的碱基组成与非编码区的差异。对于原核生物,其基因结构较为简单,一些研究小组已经在这方面取得了很好的结果,目前预测准确率可达到98%左右。对于真核生物,它们的基因结构远比原核生物的复杂,因此,要从具有较多内含子的真核生物基因组序列中正确识别出编码区是个相当困难的问题,仍然有大量的工作要做。DNA序列蛋白编码区的一个显著的特征就是1/3周期性——在DNA序列的频谱中,如果在f =1/3处存在一个尖峰,这就可能对应于三联密码子。这种周期性在蛋白编码区序列中是普遍存在的,而在大多数非编码序列中却是不存在的。因此,基因组序列的傅立叶频谱中在f=1/3处是否存在峰值,可作为判断该序列是否是编码序列的一个简单可行的标准。本文首先利用傅立叶技术分析DNA序列的三周期性,并提出一种快速计算方法。但是由于傅立叶技术分析核苷酸序列不可避免的带来随机涨落,会在频域内引入高频噪声,因此仅仅根据傅立叶分析的结果还难于达到准确识别基因的目的。小波变换是一种新的变换分析方法,具有多分辩分析的特点,而且在时频两域都有表征信号局部特征的能力,被誉为分析信号的显微镜。小波变换在一定的滤波尺度下可有效地除去随机涨落引起的高频噪声,所以本文提出以傅立叶变换为基础,再利用小波变换将随机涨落和有用信号分开,建立一种基因识别的新方法,以达到简便、快速,准确识别基因组序列编码区的目的。通过对来自于Genbank数据库的DNA序列以及标准DNA序列集ALLSEQ进行大量实验证明,基于小波变换的方法对DNA序列编码区进行预测可以取得良好效果,探测率可达到81%,优于其他方法,正确率可达到75%。新方法十分容易应用,不需要基因组序列的任何先验知识;运算十分快速,运算速度比基于快速傅立叶变换的预测方法还要提高很多倍;同时,该方法可实现对编码区的初步定位,为进一步用实验方法精更多还原

【Abstract】 In recent years, Genome projects have given rise to an exponentially growing amount of genetic information. How to find out useful information in the huge amounts of data is the problem that scientists focus on in current and future. One of the most important and basic problems is the gene identification, namely the identification of protein-coding regions in DNA sequences through computational means. In present, a number of methods for gene detection, based on distinctive features of protein coding sequences have been proposed. For example, neural net-based method, the method based on correlation function, Fourier-based analysis, and so on. At the same time, comprehensive evaluation of various methods suggest that they can’t work equally well for all genes, and constant refinement is needed to evolve better methodologies, there is also a need for new method of gene predication.For most of DNA sequence, one of the principle feature is that the dominant signal in coding regions of genomic sequences exhibits a 1/3 periodicity which is evidenced as a sharp peak at frequency f=1/3 in the power spectrum. Such periodicity is universal for protein-coding sequences and is absent in genomic sequences that do not code for proteins thus making this parameter a convenient criterion for recognizing coding sequences in DNA sequence. However, statistical analysis by Fourier technology for DNA sequence will bring noise inevitably, only according to the result by Fourier analysis for DNA sequences, it is difficult to assess the probable genes of DNA sequence exactly. Wavelet transform is a space-scale analysis. It is called as a mathematical microscope for analysis signal. The wavelet transform can eliminate the noise in a certain scale, namely, separate the useful signal from noise signal. Thereby, we can utilize wavelet transform to analyze the result by Fourier analysis for DNA sequences, and can predicate the protein-coding region conveniently, and develop the prediction accuracy.In this paper, we analyze 1/3 periodicity of protein coding region using wavelet transformation, and the result from theoretical analysis and experiments show that wavelet transform is a useful tool for detecting the periodicity. Thereby, we develop a new method to recognize the probable genes of genomic sequence. This method is更多还原