中文摘要：

本文基于图论（graph theory）思想挖掘功能相似蛋白质。将蛋白质相互作用网络（protein-protein interaction network,PPI）转化为图,使用1-hop方法把相互作用网络拆分为子图。应用最短路径法,依次把拆分得到的每一个子图与所有子图进行两两比对,寻找最优比对结果,挖掘物种内有相似功能的蛋白质。应用该方法对构建的人类蛋白质相互作用网络进行比对,共挖掘到1081对功能相似蛋白质。该方法的特点是数据量偏小、较可靠。计算结果中还出现了一些序列相似度较低、用序列比对的方法难以挖掘到的功能相似蛋白质。通过对结果的分析可以发现,趋同进化和同源有时很难区分。本实验的全部数据、程序和结果可在网站http：//bina.biositemap.com上查询。

英文摘要：

Function-similar proteins could be detected based on graph theory.Regarding PPI network as graph,and splitting the graph into subgraphs with the method of 1-hop.Working on each subgraph at a time,proteins with functional similarity in a species can be detected by aligning this gotten subgraph against each one of all the subgraphs with the method of shortest path and extracting optimal results.Aligning the built human PPI network with this method,1081 pairs of functional similarity proteins were detected.The statistical analysis of these results revealed that these detected data are small in quantity,but credible.Some proteins with low sequence similarity that makes it hard to be detected by sequence alignment could be extracted here.By analyzing these results,it＇s obvious that,sometimes,it＇s a hard work to separate homologous from convergent evolution.All the data,sotfware and results of this work are available on the website http：//bina.biositemap.com.