细菌特征分析的革兰氏阴阳性判别算法  

Gram-negative and Gram-positive Discriminating Algorithm for Bacterial Characteristic Analysis

作　　者：袁健[1] 赵桦 张明 张劲松 YUAN Jian;ZHAO Hua;ZHANG Ming;ZHANG Jing-song(Schoolof Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China;Teaching and Research Support Center of Naval Medical University,Shanghai 200433,China;Shanghai Institutes for Biological Sciences,Chinese Academy of Sciences,Shanghai 200031,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093 [2]海军军医大学教研保障中心,上海200433 [3]中国科学院上海生命科学研究院,上海200031

出　　处：《小型微型计算机系统》2021年第4期678-684,共7页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61775139,61602460,11701379)资助。

摘　　要：自然界中细菌无处不在,细菌的革兰氏阳性和阴性的有效分类对于临床治疗具有重要意义.现有的细菌的革兰氏阴阳性分类主要依赖于革兰氏染色法.这种方法借助细菌细胞壁结构的不同引起的染色性的差异来进行分类,然而涂片的厚薄和脱色时间的掌握制约着革兰氏染色法的准确性,并且实验需要花费一定时间.本文提出一种用计算机智能分析的细菌革兰氏阴阳性判别方法—基于蛋白质序列特征分析的细菌革兰氏阴阳性判别算法GCBPS.该算法首先挖掘出闭合邻接序列模式(FCloConSP)集合并对大量已知阴阳性的细菌蛋白质序列特征进行提取,然后先利用赋参的余弦相似度距离计算方法来衡量待测细菌蛋白质序列与阳性细菌特征集之间的距离来初步判别是否为阳性,再通过去假阴性等处理后得到最终的细菌革兰氏阴阳性判别结果.该算法已在标注的1591条革兰氏阴性菌以及576条革兰阳性菌的标准数据集上进行评估,实验结果表明,判别的平均正确率F1值可达到95.4%.Bacteria are ubiquitous in nature,and effective classification of Gram-positive and Gram-negative bacteria is important for clinical treatment. Gram-negative and Gram-positive classification of existing bacteria mainly relies on Gram staining. This method uses the differences in staining caused by the structure of the bacterial cell wall to classify. However,the thickness of the smear and the grasp of the decolorization time limit the accuracy of the Gram staining method,and the experiment takes some time. This paper presents a method for distinguishing gram-negative and gram-positive bacteria by computer intelligence analysis-GCBPS,a gram-negative and gram-positive discrimination algorithm for bacteria based on protein sequence feature analysis. The algorithm first mines a closed adjacency sequence pattern( FCloConSP) set and extracts a large number of known Gram-negative and Gram-positive bacterial protein sequence features. Then it uses the cosine similarity distance calculation method with parameters to measure the distance between the bacterial protein sequence and the Gram-positive bacteria feature set to determine whether it is Gram-positive,and then remove the false Gram-negative treatment to obtain the final bacteria Gram-negative and Gram-positive Judgment result. The algorithm has been evaluated on a standard data set of labeled 1591 gram-negative bacteria and 576 gram-positive bacteria. The experimental results show that the average accuracy rate of discrimination F1 can reach 95.4%.

关 键 词：革兰氏阴阳性判别 蛋白质序列分析 闭合序列模式 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]