Using of Normalizations for Gene Expression Analysis

Normalizations of gene expression data are commonly used in practice. They are used for removing systematic variation which affects the measure of gene expression levels. But one can object to the using of normalized data for testing hypotheses. By using normalized data, tests can break nominal level of multiple testing on which we would like to test the hypotheses. It could bring a lot of false positives, which we would like to prevent. In this chapter, by simulating data with similar correlation structure as real data, we try to find out how quantile, global, and δ -sequence normalizations hold the nominal level of Bonferroni multiple testing procedure.