摘要
提出了一种基于偏最小二乘判别分析和F-score的特征筛选方法,并将其用于蛋白质组学质谱数据分析。方法主要包含3个步骤:(1)用LIMPIC算法对原始数据进行预处理;(2)计算每个变量的F-score值并将所有变量按F-score值降底的顺序排列;(3)采用偏最小二乘判别分析交互检验按前向选择法选择最佳变量子集。用本方法对一组卵巢癌数据进行分析,最终从原始的15154个质荷比变量中选择了12个特征变量作为潜在生物标记物,它们在训练集上交叉检验的特异性和灵敏度分别为98.36%和98.15%,在独立测试集上的特异性和灵敏度分别为96.67%和100%。用筛选出的变量作PCA所得的结果显示这些变量能够较好地将样本分类,说明能够反映出样本的类别信息。所提出的方法可用于蛋白质组学质谱数据的特征筛选及样本分类。
A feature selection method based on F-score and partial least square discriminant analysis (PLS-DA) was proposed and used for proteomic mass spectrometric (MS) data analysis and potential biomarker discovery. The method mainly includes 3 steps : ( 1 ) spectra preprocessing with LIMPIC algorithm; (2) calculating the F- score values for each variable and sorting them according to their F-score values in descending order; and (3) determination of the optimum feature set with PLS-DA cross validation in a forward stepwise selection manner. An ovarian cancer dataset was analyzed with the proposed method. As results, 12 m/z locations were selected as potential biomarkers. The features could distinguish the disease samples from healthy controls on the independent test sets with sensitivity of 100% and specificity of 96. 67%. The results show that the method proposed in this study is available for classification feature selection from proteomic MS data.
出处
《分析试验室》
CAS
CSCD
北大核心
2012年第10期106-109,共4页
Chinese Journal of Analysis Laboratory