Mining Coregulated Biclusters From Time-series Gene Expression Data

The DNA microarray is an important technique in biology, which can monitor the expression levels of thousands of genes. Through analyzing the data of microarry, Biologist can get significant predictions on gene and gene regulation network.Clustering, especially biclustering, is widely used to analyze microarray data. Most traditional algorithms focus on finding clusters in which genes have similar changing trends under a series of time points. However, in most of the biological applications, gene clusters with the similar and opposite changing trends could deliver more valuable information,Hence, in this thesis, we introduced the concept of co-regulated clusters, based on which we proposed an efficient algorithm to mine genes with positive and negative regulations (same and opposite changing trend) from 2D microarray gene expression data. Moreover, we have extended the 2D mining algorithm to solve high dimensional3D gene expression. Experimental results with real microarray yeast data show that our algorithms could effectively deliver significant co-regulated clusters for biological study.