蛋白质序列中的关联规则发现及其应用  

DISCOVERY AND APPLICATIONS OF THE ASSOCIATION RULES IN PROTEIN SEQUENCES

作　　者：陈双平[1] 郑浩然[2] 刘海燕[3] 王煦法[2] 

机构地区：[1]中国科学技术大学电子工程和信息科学系,合肥230027 [2]中国科学技术大学计算机科学与技术系,合肥230027 [3]中国科学技术大学生命科学学院,合肥230027

出　　处：《生物物理学报》2006年第3期171-176,共6页Acta Biophysica Sinica

基　　金：中国科学院知识创新工程重要方向项目(KSCX2-SW-329)~~

摘　　要：随着蛋白质序列-结构分析中使用的机器学习算法越来越复杂,其结果的解释和发现过程也随之复杂化,因此有必要寻找简单且理论上可靠的方法。通过引入原理简单、理论可靠、结果具有很强实际意义的关联规则发现算法,找到了蛋白质序列中数以万计的模式。结合实例演示了如何将这些模式应用于蛋白质序列分析中,如保守区域发现、二级结构预测等。同时根据这些结果构建了一个二级结构规则库和一种简单的二级结构预测算法,实验结果表明,约81%的二级结构可以由至少一条关联规则预测得到。In recent years, many complicated motif-discovering algorithms have been developed to find the motifs in protein sequences, but their results and running processes are hard to be interpreted. Association rules discovery is a simple method, but it has a theoretical background of probability. Applying the method, thousands of motifs have been found, which can be used in protein sequence analysis such as preserved site discovery and secondary structure prediction. A secondary structure rule library is constructed using the found association rules, and a simple secondary structure prediction algorithm is developed. It shows that nearly 81 percent of secondary structure can be implied by at least one rule.

关 键 词：蛋白质序列 关联规则 二级结构预测 模式发现 

分 类 号：Q518.1[生物学—生物化学] Q71