Exploiting biclustering for missing value estimation in DNA microarray data

The missing values in gene expression data harden subsequent analysis such as biclustering which aims to find a set of coexpressed genes across a number of experimental conditions. Missing values are thus required to be estimated before biclusters detection. Existing estimation algorithms rely on finding coherence among expression values throughout the entire genes and/or across all the conditions. In view that both missing values estimation and biclusters detection aim at exploiting coherence inside the expression data, we propose to integrate them into a single framework. The benefits are twofold, the missing value estimation can improve bicluster analysis and the coherence in detected biclusters can be exploited for better missing value estimation. Experimental results show that the integrated framework outperforms existing missing values estimation algorithms. It reduces error in missing value estimation and facilitates the detection of biologically meaningful biclusters.