中文摘要：

图像配准是图像处理的一个重要技术,可用于分析两幅图像之间的相似度.本文提出了一种基于图像配准分析物种进化关系的新方法：首先利用一阶马尔可夫链方法计算不同基因组序列的寡聚核苷酸转移概率矩阵;然后将转移概率矩阵转换为彩色图像矩阵,并绘制物种两两之间彩色图像矩阵的联合直方图;最后分析联合直方图点集的分布情况,引入直方图点集的散度公式,将其作为相似性测度的标准,从而鉴定物种亲缘关系的远近.100种细菌全基因组的计算结果表明,相较于单基因法或基于基因组寡聚核苷酸频率组分差异信息的方法,本文提出的新方法具有更高的准确度和分辨力,它不仅能够很好地分辨科以下的分类单元,而且对科以上的分类单元同样具有较好的区分效果.该方法有望发展成为物种鉴定及系统发育推断的有效手段.

英文摘要：

Image registration is an important technique in image processing, which could be used to analyze the similarity between two images. Here, a novel method based on image is proposed to infer the evolutionary relatedness of various microbial organisms. Firstly, the oligonucleotide transition probability matrices of microbial genomes were calculated by applying 1st order Markov Chain Method. Secondly, each transition probability matrix was converted into a color image, and then combined with each other to depict as a joint histogram. Finally, the point set distribution of joint histogram was analyzed, and divergence formula was brought in and used as the similarity reference standard to determine the evolutionary relatedness of organisms. According to the study results of 100 bacterial genomes, our results suggest that this new method is more accurate and discriminable than the methods based on single gene or genomic oligonucleotide-based frequency component difference information methods. It not only can distinguish microbial organism under the family of taxonomy, but also has a better capability to distinguish microbial organism beyond the family of taxonomy. This method is expected to develop into effective device and apply to species identification and phylogeny inferring.