Improving Missing Value Imputation in Microarray Data by Using Gene Regulatory Information

Accurate estimation of missing values in microarray data is important for the expression profile analysis. In this paper, missing value imputation is done with the aid of gene regulatory mechanism. It incorporates histone acetylation into the conventional k-nearest neighbor and local least square imputation algorithms for final prediction. The comparison results indicated that the proposed method consistently improves the widely used methods and outperforms GOimpute in terms of normalized root mean squared error(NRMSE), which is one of the existing related methods that use the functional similarity as the external information. The results demonstrated histone acetylation information may be more highly correlated with the gene expression than that of functional similarity.