Missing value estimation of microarray data using similarity measurement

Abstract

DNA gene expression profiling plays an important role in a wide range of areas in biological science for handling cancer diseases. Data generated in microarray related experiments have many missing expression values which lose valuable information from the dataset. The proposed method first partitions the genes without missing values using clustering algorithm and then measures the similarity between a gene with missing values and the centroid of the clusters and finally, the missing values are estimated by the corresponding expression values of the centroid giving maximum similarity factor. The method explicitly depends on expression values to imputes missing values, completed the input dataset with low errors for data analysis and knowledge discovery. The method is compared with prominent approaches, such as zero-impute, row-average-impute and KNN-impute in terms of "Normalized Root Mean Square Error" to claim its novelty. © 2012 Springer-Verlag.