摘要
从蛋白质的一级序列出发,用矩阵打分的方法对3088个蛋白质中的β发夹和非β发夹模体进行了识别.使用10-交叉检验,预测总精度为75.9%,Matthew相关系数为0.42.同时计算了不同loop长的模体对应的序列最佳固定模式长,并对有相同最佳固定模式长的模体序列进行了组合,组合后的模体预测总精度都高于76.1%,Matthew相关系数大于0.43.
Based on the protein sequence,the β-hairpins and non fl-hairpins in the 3088 proteins are recognized by using the scoring matrix. The overall accuracy of prediction and Matthew's correlation coefficient are 75.9% and 0. 42,respectively,with 10-fold cross-validation. Moreover,the best fixed length pattern for the different loop lengths are given. By using of the combination patterns with the same fixed length pattern, the total accuracy of prediction and Matthew's correlation coefficient are higher than 76.1% and 0. 43 respectively.
出处
《内蒙古大学学报(自然科学版)》
CAS
CSCD
北大核心
2007年第6期654-659,共6页
Journal of Inner Mongolia University:Natural Science Edition
基金
国家自然科学基金资助项目(30560039)
内蒙自然科学基金资助项目(200508010509
200607010101)
关键词
Β-发夹模体
位置频率矩阵
位点保守性参量
打分函数
β-hairpin motif position probability matrix
conservation index vector of position scoring function