SBi-MSREimpute: A Sequential Biclustering Technique Based on Mean Squared Residue and Euclidean Distance to Predict Missing Values in Microarray Gene Expression Data

Abstract

Due to technical problems in DNA microarray technology, the output matrix named gene expression data contains a huge number of missing entries. These missing entries create problems when analysis algorithms like classification, clustering, etc. are applied on microarray gene expression data as these methods require complete data matrix. To solve the above-mentioned problem, several missing value imputation techniques have been developed. Among them clustering- and biclustering-based missing value prediction methods are most popular due to their simplicity. In this regard, here a new biclustering-based sequential imputation method is proposed. In this method, for every missing position, a bicluster is formed in a novel manner using the concept of mean squared residue (MSR) score and Euclidean distance. Then the imputation is carried out sequentially by computing the weighted average of the neighbour genes and samples present in the bicluster. To evaluate the performance, the proposed method is rigorously tested and compared with the well-known existing methods. The effectiveness of the proposed method, is demonstrated on different microarray data sets including time series, non-time series, and mixed.