基于粗糙集的蛋白质结构分类属性筛选  

Selection of protein structure classification characteristics based on rough set

作　　者：敖培[1] 王川[1] 张纪[1] 

机构地区：[1]河南师范大学计算机与信息工程学院,河南新乡453007

出　　处：《计算机应用研究》2014年第5期1328-1331,共4页Application Research of Computers

基　　金：河南省教育厅科学技术研究重点项目基础研究计划资助项目(13A413506);河南省教育厅科学技术研究重点项目(12A520028);河南师范大学新引进博士科研启动费支持课题(qd12136);河南省软科学研究计划项目(112400450305)

摘　　要：为了对蛋白质结构进行正确分类,提出了一种基于粗糙集理论的蛋白质结构分类属性筛选方法。通过多结构比对工具MAMMOTH-mult获得条件属性值,针对分辨矩阵中元素特点提出了分辨矩阵简化方法和改进的属性约简方法。实验以SCOP 1.71数据库中结构信息完整的35个家族数据集为研究对象,采用本方法得到%STRCTCORE和%LOOSECORE两个蛋白质分类属性,并通过两个属性的d1a0fa1与35个蛋白质家族、46 626家族与35个结构比对结果散点图可以看出,将这两个分类属性作为蛋白质结构分类标准,基本上可以对蛋白质结构进行客观正确的分类。Scientific and reasonable classification attribute is critical for proper protein structure classification. This paper presented a method of selecting protein structure classification attributes based on rough set theory. Attribute values were ob- tained by multiple structure alignment tool MAMMOTH-mult. It proposed methods of simplified discernibility matrix and im- proved attribute reduction according to characteristic of elements in discernibility matrix. Data set of 35 families with complete information was taken as the research object in SCOP 1.71 database. The proposed method obtained % STRCTCORE and % LOOSECORE protein classification attributes. From scatter diagrams of structure alignment between dl a0fal and 35 fami- lies,46 626 family and 35 structures, it shows that the two protein classification attributes are taken as the standard of protein structure classification, protein structure can be classified correctly.

关 键 词：粗糙集 分辨矩阵 属性约简 多结构比对 蛋白质结构分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]