【作者】 卢颖；

【导师】 潘泉； 张洪才；

【作者基本信息】 西北工业大学 ， 控制理论与控制工程， 2003， 硕士

【摘要】 进入后基因组时代，随着生物信息学的发展和基因组数据的飞速积累，基因功能的研究逐渐成为重点。这就需要将生物序列作为一个有机的整体，来考察其在系统层次上交现出来的整体特征。本文从全局和细节两个角度对DNA全序列的分形特征和自相似特性进行了研究，以求更全面更准确地反映DNA序列的分形现象。 1．在DNA序列的可视化研究中，利用语言学的思想和可视化框图法，从全局的角度在二维Portrait图上展示了完整基因组序列的分形特征，并对其加以分析，发现其分形特征与序列中各碱基串的分布和含量有密不可分的关系。 2．在对DNA序列分形特征作进一步研究方面，首先用小波变换和分段平均傅立叶频谱法对其自相似性和长程相关性进行检验，紧接着用R／S分析法和基于小波的间接估计方法对其分形参数即Hurst系数进行了估计和比较分析，发现在不同尺度下Hurst系数不一致，并且针对Portrait图所表现出的明显的分形特征作了进一步研究，把一维小波变换推广到二维，为二维信号的分形参数估计提供了一种较为合理的参考。结果表明，DNA在全序列尺度上表现出渐进自相似结构，只有在低频部分才可以用严格自相似信号来描述。 3．为便利后续工作，总结成果，设计了一个针对DNA序列分形特征研究的软件。该软件有友好的用户界面，具有一定的使用价值。 本论文的研究工作可以定性地解释DNA序列的分形现象。同时，在总结工作的基础上提出了一些新的问题和设想，希望对今后的生物学研究会有一定的参考价值。更多还原

【Abstract】 In the post-genomic era, with the development of informatics and the progress of genome sequencing projects, researches are gradually focused on the function of genes. Organisms are being viewed as an integrated system. The properties emerged in systematic levels are being studied. In this thesis, the fractals of DNA sequences, including the self-similarity structure, are researched with two aspects of full-scale and details, in order to show the fractals of DNA sequences more globally and more exactly.1. In visualizing long DNA sequences, including the complete genomes of several bacteria, language method and the visualization scheme are introduced to express and discuss globally the fractal-like pattern of the complete genomes in the figure of portrait. As a result, the close relation between the fractal-like patterns and the difference in the g, c, a and t contents in the sequences are uncovered.2. The wavelet transform and the segment average Fourier frequency-spectrum are use to inspect the self-similarity. The two parametric methods are proposed to characterize the self-similarity properties, and mainly used to estimate and analyze the Hurst coefficient. And the variance of Hurst coefficient on the different scales is discovered. Also, the visible fractals of portrait figure are studied by extending the one dimension wavelet transform into two dimensions and this will provide a relatively reasonable reference to the parameter estimation of fractals in two dimensions. Results show that long-range correlation structure prevails through the entire molecule of DNA, and the strict self-similarity is set up to describe DNA sequences only on the segment of low frequency.3. The software with friendly user interface is designed to make the future work more convenient.The work in this thesis can explain qualitatively the fractals of DNA sequences. At the same time, some new problems and suggestions can be inferred from the research, which can be helpful to future bioinformatics studies.更多还原