Differentially Expressed Genes in Microarray Experiment with Dye-swapping Design

Abstract

Identifying differentially expressed genes is one of the major goals in microarray analysis. To address inherent variations, various methods and experimental designs have been proposed. In the present study, we present an easily implemented statistical method, SARA (Swapping And Regression Analysis), which is special tailored for microarray experiments that utilize dye swapping and replications. Under very general assumptions regarding the structure of channel, scanner, and target effects obtained from the experiment, we proved that SARA can efficiently removes bias caused by the above effects. Our methods and a few commonly used methods including ANOVA (Kerr et al. 2000), CAT (Tsodikov et al. 2002), fold-change and t-test type methods were compared using simulation and real data. Our statistical framework and simulation results show that SARA has satisfactory performance. Programs implementing the proposed method is available at http://www.ibms.sinica.edu.tw/~csjfann/first%20flow/program.htm