一种无参数的微阵列缺失值填补方法(英文)  

A non-parametric imputation method for microarray missing values

作　　者：张小白[1] 王惠南[1] 宋晓峰[1] 张焕萍[1] 

机构地区：[1]南京航空航天大学自动化学院生物医学工程系,江苏南京210016

出　　处：《计算机与应用化学》2007年第12期1611-1616,共6页Computers and Applied Chemistry

摘　　要：微阵列数据中的缺失值会对随后的数据分析造成影响。因此,正确地估计这些缺失值是很必要的。将一个k值选取算法结合到有序的局部最小二乘填补算法中,提出了一种无参数的缺失值填补方法(SLLSkimpute)。该方法的三个特点是:第一,无需事先确定参数;第二,针对不同的目标基因使用不同数目的邻居基因;第三,有序地估计缺失值,并有选择地将已得到的估计值应用到后续的估计过程中。实验结果证实了该算法的有效性,其估计性能优于其它一些常用的填补方法。Missing values contained in microarray data will affect subsequent analysis. It is thus essential to estimate these missing values accurately. Incorporating a k-value selection algorithm into sequential local least squares imputation, a non-parametric imputation method called SLLSk impute is proposed to estimate missing values. Three unique features of SLLSk impute are: firstly, no parameter is required beforehand; secondly, different numbers of neighboring genes are used for different target genes; thirdly, missing values are estimated sequentially and the imputed values are selectively used for the following imputation. Experimental results confirmed that SLLSk impute method is valid and it exhibited better estimation ability than other imputation methods used currently.

关 键 词：缺失值 填补方法 最小二乘法 标准均方根误差(NRMSE) 微阵列数据 

分 类 号：Q332[生物学—遗传学]