基于拓扑和语义相似性的疾病基因识别算法-计算机技术专业论文.docxVIP

摘要 疾病一直以来威胁着人类的健康，虽然当今社会拥有更先进的医疗设备和诊 断方法，但是，疾病问题依然很严峻。随着人类基因组全序列测序与工作草图的 完成，生命科学进入了后基因组时代，生物科学技术得到迅猛发展，数据资源也 快速增长。因此，采用计算的方法从大量的生物数据中挖掘疾病和基因之间的关 系得到了越来越广泛的关注。 研究表明在蛋白质相互作用网络中有相互作用关系的编码该蛋白质的基因或 者有相同功能的基因往往会导致相同的或者相似的疾病，据此提出了很多基于蛋 白质相互作用网络的疾病基因识别方法。但是，通过分析，一些导致相同疾病的 基因在网络中表现出较弱的相互作用关系，从而影响疾病基因的识别工作。因此， 本文提出了整合蛋白质相互作用网络的拓扑结构相似性和基因之间的语义相似性 的方法来提高疾病基因识别的准确性。首先计算蛋白质相互作用网络节点之间的 拓扑结构相似性，并以所有编码该网络中蛋白质的基因作为候选疾病基因，其次 计算蛋白质网络中所有候选基因的基于生物学路径的语义相似性，最后将这两种 相似性进行整合，从而对候选基因进行得分排序。在此工作的基础上，选择排名 靠前的候选疾病基因应用到疾病模块识别中，通过查询基因本体论数据库对候选 模块进行富集性分析。 在实验部分，将本文方法与重启的随机游走进行比较，其平均排名比提高了 4.57%，同时在具体的实例中进行验证和分析，证明了本文方法的有效性和可靠性。 关键词：疾病基因识别 拓扑相似性 语义相似性 蛋白质相互作用网络 Abstract Diseases have always been threatening human health. Although in present society we have more advanced medical equipments and diagnostic methods, disease problem is still very serious. Along with the completion of the human genome sequencing and the draft work, life science has entered into the post-genomic era. Biological science and technology have been developing rapidly and biological data resources also get rapid growth. Therefore, methods based on computer science to mine the relationships between diseases and genes from a large number of biological data get more and more widely attention. Lots of research show that genes which encode proteins that interact with each other in the protein-protein interaction network or with the same functions often lead to the same or similar disease. With this assumption, bioinformatics researchers have put forward many disease gene identification methods based on protein-protein interaction network. However, through our analysis, we found that some genes that lead to the same or similar disease in the protein-protein interaction network show weak interaction relations, which restrict the disease gene identification work. On this basis, we put forward a method that integrates topology structure similarity in protein-protein interaction network and semantic similarity between genes to improve di