基于离散平稳小波分解和递归零空间LDA算法的SELDI蛋白质谱特征选择  

Feature Selection for SELDI Mass Spectrometry Data Based on Discrete Stationary Wavelet Transform and Null Space LDA Algorithm

作　　者：祝磊[1] 王尧佳[1] 韩斌[1] 厉力华[1] 柯激情[1] 孟旭莉[2] 

机构地区：[1]杭州电子科技大学生命信息与仪器工程学院,浙江杭州310018 [2]浙江省肿瘤研究所,浙江杭州310022

出　　处：《航天医学与医学工程》2012年第4期260-265,共6页Space Medicine & Medical Engineering

基　　金：国家自然科学基金(60801054;60801055);浙江省公益性技术应用研究项目(2010C33017);浙江省医药卫生科学研究基金(2010KYA041)

摘　　要：目的针对如何筛选与肿瘤相关的蛋白位点问题,提出一种基于平稳小波变换与递归零空间LDA算法相结合的特征选择方法。方法首先对样本质谱数据进行平稳小波变换;接着基于递归框架调用零空间LDA算法,挑选出最具判别意义的小波系数特征;然后经平稳小波逆变换,将挑选出的小波系数特征对应回原始蛋白质谱数据中,获得与肿瘤判别相关的蛋白位点。最后,运用SVM分类器估算位点的分类性能。结果在卵巢癌公共数据集OC-WCX2b和浙江省肿瘤医院乳腺癌数据集BC-WCX2a上分别挑选出与肿瘤判别相关的6个和2个蛋白位点。结论本文提出的算法能够有效提取出具有较好判别效果的蛋白质谱位点,有助于癌症的辅助诊断。Objective To propose a new feature selection algorithm based on Discrete Stationary Wavelet Trans- form （DSWT） and recursive null space LDA （NLDA） for selection of the cancer related protein feature from the mass spectrometry data. Methods Firstly, the DSWT was performed on sample mass spectrometry data. Then the NLDA algorithm was adopted to select the most discriminative wavelet features. Thirdly, the Inverse Stationary Wavelet Transform was used to find the original protein features which were corresponding with the wavelet features. Finally the SVM classifier was used to estimate the performance of the algorithm. Results The proposed method was tested and evaluated on the ovarian cancer database OC-WCX2b, and breast cancer database BC-WCX2a and 6 m/z and 2 m/z features were selected respectively. Conclusion The experimental results demonstrate a good performance of the proposed method which will be more helpful for cancer diagno- S1S

关 键 词：蛋白质质谱 特征选择 离散平稳小波变换 零空间LDA 

分 类 号：Q789[生物学—分子生物学]