Clustering of short time-course gene expression data with dissimilar replicates

Microarrays are used in genetics and medicine to examine large numbers of genes simultaneously through their expression levels under any condition such as a disease of interest. The information from these experiments can be enriched by following the expression levels through time and biological replicates. The purpose of this study is to propose an algorithm which clusters the genes with respect to the similarities between their behaviors through time. The algorithm is also aimed at highlighting the genes which show different behaviors between the replicates and separating the constant genes that keep their baseline expression levels throughout the study. Finally, we aim to feature cluster validation techniques to suggest a sensible number of clusters when it is not known a priori. The illustrations show that the proposed algorithm in this study offers a fast approach to clustering the genes with respect to their behavior similarities, and also separates the constant genes and the genes with dissimilar replicates without any need for pre-processing. Moreover, it is also successful at suggesting the correct number of clusters when that is not known.