联合稀疏独立法则及其在疾病分类中的应用  

JOINT SPARSE INDEPENDENCE RULE AND ITS APPLICATION IN GENE EXPRESSION ANALYSIS

作　　者：王丽平[1] 吴亚飞[1] 

机构地区：[1]南京航空航天大学理学院,南京100016

出　　处：《数值计算与计算机应用》2016年第3期179-185,共7页Journal on Numerical Methods and Computer Applications

基　　金：国家自然科学基金(11471159;6166136001);江苏省基础研究计划项目(BK20141409)资助

摘　　要：通过有效的特征选择进行疾病分类是当前生物信息学研究的常见问题,从高维基因数据中消除噪声、筛选出存在于低维子空间的特征基因,对预测、诊断和治疗疾病至关重要.基于新兴的联合稀疏优化算法与经典的独立法则,本文提出了联合稀疏独立分类方法.在特征选择时考虑了数据的整体稀疏结构及集体特征之间的关系,弥补了基因表达分析数据小样本信息不足的缺点.而利用独立法则进行疾病分类不仅方式简单、易于实现,并且有效阻断了噪声的相互干扰,具有较好的稳定性.在三个基因表达分析数据集上的疾病分类实验结果表明,新的分类方法具有良好的分类正确率和运行速度.Disease classification by effective feature selection is a common problem in bioinformat- ics research. Selecting the most important genes from high dimensional genetic data and eliminating noise are essential for the prediction, diagnosis and treatment of diseases. Based on the recently-developed joint sparse algorithms and the classical independence rule, this paper proposes a joint sparse independence classification rule. The new scheme considers the integral sparse structure and the correlation between collective features for selecting features which makes up the weakness in small sample data. Independence rule for disease classifi- cation is not only economical and easy to implementation but also effective to prevent noise from accumulating. The experimental results on three gene expression databases confirm the good performance of the new classification methn~ in ~ A ;~ ＊;--^

关 键 词：联合稀疏优化 独立法则 小样本数据 基因表达 

分 类 号：R366[医药卫生—病理学]