Comparison of Normalization Methods in Microarray Analysis

Abstract

<p> DNA microarrays can measure the gene expression of thousands of genes at a time to identify differentially expressed genes. The Affymetrix GeneChip system is a platform for the high-density oligonucleotide microarray to measure gene expression using hundreds of thousands of 25-mer oligonucleotide probes.</p> <p> To deal with Affymetrix microarray data, there are three stages of preprocessing to produce gene expression measurements/values. These are background correction, normalization and summarization. At each stage, numerous methods have been developed.</p> <p> Our study is based on Affymetrix MG_U74Av2 chip with 12488 probe sets. Two strains of mice called NOR and NOR.NOD_Idd4/11 mouse are hybridized for the experiment. We apply a number of commonly used and state-of-art normalization methods to the data set, thus compute the expression measurements for different methods. The major methods we discuss include Robust Multi-chip Average (RMA), MAS 5.0, GCRMA, PLIER and dChip.</p> <p> Comparisons in terms of correlation coefficient, pairwise expression measures plot, fold change and Significance Analysis of Microarray (SAM) are conducted.</p>