Rough fuzzy cuckoo search for triclustering microarray gene expression data

Analyzing time series microarray gene expression data is a computational challenge due to its three-dimensional characteristics. Triclustering techniques are applied to three-dimensional data for mining similarly expressed genes under a subset of conditions and time points. In this work, a novel rough fuzzy cuckoo search algorithm is proposed for triclustering genes across samples and time points simultaneously. By applying the upper and lower approximation of rough set theory and the objective function of fuzzy k-means, rough fuzzy k-means was incorporated into a cuckoo search to handle the uncertainty of the data. The proposed method was applied to three real-life time series gene expression datasets. This work was evaluated using four validation indices and correlation analysis was performed to indicate the cluster quality. The proposed work was also compared with the existing triclustering algorithms and it outperformed the other methods.