An introduction to gene expression data clustering algorithims
Guo, Ling Qiong
Date of Issue2008
School of Physical and Mathematical Sciences
DNA microarray, also called oligonucleotide array, gene array, DNA chip, gene microarray, or genome chip, is a high-throughput technology that allows detection of expression of thousands of genes simultaneously. Among a variety of its experimental protocols, one is to first fix thousands of DNA probes on an objective buttress by situ-synthesis or micro-print to produce a two-dimension DNA probe matrix, and then hybridize these DNA probes with a biological (cDNA or cRNA) sample. Microarray expression data is hence collected by measuring the relative abundances of the hybridized probes, to quantify the activities of genes under a biological experiment. Therefore, as commonly believed, the biological signals of genes are hidden in microarray expression data. However, detecting these hidden biological signals presents a big challenge to the microarray research community. Cluster analysis is one of analytical techniques to detect the biological signals hidden in microarray expression data. Given an expression dataset, it aims to find groups of genes whose expression profiles are similar to each other within a group while dissimilar between groups. Clustering often serves as a very preliminary step in gene expression analysis, and allows us to gain valuable insights into the underlying biological mechanism. Clustering is a hard while well-studied problem. There are many algorithms that have been developed in the literature, some of which are intended in particular for gene expression data clustering. The main goal of this thesis is to discuss many popular clustering algorithms which have been extensively applied to microarray expression data for biological discovery, including hierarchical clustering, K-means clustering, SOM-based clustering and model-based clustering.