一种无参数的微阵列缺失值填补方法

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。

A non-parametric imputation method for microarray missing values

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.