Normalization for two-channel microarray data.

Abstract

OBJECTIVES: In two-channel microarray experiments the measured gene expression levels are affected by many sources of systematic variation. Normalization refers to the process of removing such systematic sources of variation, to make measured intensities within and between slides comparable. Some commonly used normalization methods removing intensity-dependent dye bias and adjusting differences in variability between slides will be reviewed with the main focus on intensity-dependent normalization methods. METHODS: This article describes different intensity-dependent within-slide normalization methods for the log ratios of red and green channel intensities but also refers to single channel normalization methods incorporating all single channels of the slides at once. RESULTS: The described procedures provide a useful approach to remove systematic sources of variation like intensity-dependent dye bias and variability between slides in cDNA microarray experiments. This is illustrated by an experimental data set. CONCLUSIONS: Several reasonable normalization procedures for two-channel microarray data have recently been proposed. Deciding on which method would perform well for a concrete experiment is difficult. Designed spike-in experiments or dilution series with known differences for some selected genes would be helpful to assess the different methods, but may be impractical for most laboratories due to the high costs.