δTRIMAX: erxtracting triclusters and analysing coregulation in time series gene expression data

In an attempt to analyse coexpression in a time series microarray gene expression dataset, we introduce here a novel, fast triclustering algorithm δ-TRIMAX that aims to find a group of genes that are coexpressed over a subset of samples across a subset of time-points. Here we defined a novel mean-squared residue score for such 3D dataset. At first it uses a greedy approach to find triclusters that have a mean-squared residue score below a threshold δ by deleting nodes from the dataset and then in the next step adds some nodes, keeping the mean squared residue score of the resultant tricluster below δ. So, the goal of our algorithm is to find large and coherent triclusters from the 3D gene expression dataset. Additionally, we have defined an affirmation score to measure the performance of our triclustering algorithm for an artificial dataset. To show biological significance of the triclusters we have conducted GO enrichment analysis. We have also performed enrichment analysis of transcription factor binding sites to establish coregulation of a group of coexpressed genes.