摘要
从蛋白质的氨基酸序列出发,采用"全体分类"的分类策略识别27类折叠子.当使用5种矩阵打分的方法识别27类折叠子时,预测结果并不理想,进一步使用了以矩阵打分方法得到的分值及氨基酸的二肽组分共同作为参数的离散增量算法,取得了较好预测效果.总的预测成功率达到了45.43%.
Based on the amino acid sequence of protein,27class protein fold patterns are recognized by using ＂ensemble classifier＂ strategy.Five scoring matrixes are use for recognition,but the result turns out to be not quite desirable.Then,for better predicting 27-class protein fold patterns,an algorithm of increment of diversity is proposed,whose parameters include not only the value of scoring matrixes but also the amino acid dipeptide composition.The overall prediction accuracy of 27 protein fold patterns reaches 45.43%.
出处
《内蒙古工业大学学报（自然科学版）》
2010年第1期18-24,共7页
Journal of Inner Mongolia University of Technology：Natural Science Edition
基金
内蒙古自治区高等学校科学研究项目(NJZY08059)
国家自然科学基金资助项目(30960090)
关键词
打分矩阵
离散增量
蛋白质折叠子预测
scoring matrix
increment of diversity
protein fold pattern prediction
