ABSTRACT
Normalization is a preprocessing procedure that is routinely applied to microarray data. The objective of normalization is to minimize, or even remove, the effects on data of artifacts that were inevitably introduced into the measurement due to imperfection of the technologies. Only after proper normalization can we contend with confidence that the identified changes in methylation or expression are of biological origins. Although nonbiological effects can be modeled together with biological ones in the analysis, we opt for a two-stage approach where methylation data are normalized first. Normalized data are then subject to exploration by various methods, which are the topics of subsequent chapters. Because of the two different types of microarrays, the ways they are normalized are somewhat different. Finally, in cases where the principles underlying the normalization are not well justified, we resort to the hybridization of known DNA fragments with the counterpart control probes on the microarray for normalization.
