一种新的蛋白质亚细胞定位预测训练集构造方法  

A novel method of training set construction for predicting protein subcellular localization

作　　者：曹隽喆[1] 顾宏[1] 贺建军[2] 

机构地区：[1]大连理工大学控制科学与工程学院,辽宁大连116024 [2]大连民族学院信息与通信工程学院,辽宁大连116602

出　　处：《大连理工大学学报》2012年第6期884-889,共6页Journal of Dalian University of Technology

基　　金：国家自然科学基金资助项目(61174027);国家科技重大专项资助项目(2010ZX04007-011-5);辽宁省自然科学基金资助项目(20102025)

摘　　要：设计了一种新的蛋白质亚细胞定位预测训练集构造方法.该方法针对传统预测方法缺乏足够的实验标记数据的问题,基于主动学习策略从非实验标记蛋白质数据中主动选择有效数据,并与原有的实验标记数据共同训练预测模型,以提高基准分类器的预测精度.结合支持向量机分类器,该方法在病毒蛋白质独立测试集上进行了预测实验,测试结果表明,该方法能够有效地提高基准分类器的预测能力,性能优于现有的病毒蛋白质预测系统.A novel method of training set construction for predicting protein subcellular localization is designed.The proposed approach aims at the shortage of valid data with experimental annotations in the traditional methods,and employs active learning strategy to actively select some effective data with non-experimental annotations which train the prediction model together with the original data,and also improves the prediction accuracy of the benchmark classifier.Combined with a support vector machine classifier,the approach is tested on an independent testing dataset of viral proteins.The test results indicate that the presented method can effectively improve the predicting capability of the benchmark classifier and can achieve a better performance than the existing predicting systems for viral proteins.

关 键 词：生物信息学 机器学习 蛋白质亚细胞定位 非实验标记数据 主动学习 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]