protocol：Normalizing Expression Arrays.docx

Normalizing Expression ArraysKey ideasNormalization is the attempt to compensate for systematic technical differences between chips, to see more clearly the systematic biological differences between samples. Biologists have long experience coping with technical variation between experimental conditions that is unrelated to the biological differences they seek. However expression arrays may vary in even more ways than measures such as rt-PCR. In practice methods that have worked well for rt-PCR and similar measures do not perform as well for microarray data,which shows many dimensions (here many dimensions means simultaneous inferences)of systematic differences. No lab technician can reproduce exactly the same technical conditions in each assay of a long series. Researchers do not have detailed data on the procedures and therefore cannot compare all the relevant technical condition of individual hybridizations.The key assumption that enables us to normalize microarrays is that only a few genes are expressed at really different levels between samples (here a few’ means up to a few hundred out of tens of thousands). Hence the expression levels of the majority of genes are essentially identical across samples. This assumption may not always be held, for example when comparing highly transformed cancer cells with normal cells. However,we need to assume something of this sort if we are to make comparisons at all, since the fluorescent intensity measures can easily be manipulated by twiddling settings on the photo-multiplier tube and these measures are routinely affected by different amounts one sample than of another and different efficiencies of label incorporation. A more specific assumption is that microarray measures should not be correlated with technical characteristics of probes, such as: base content,dye bias, intensity effects,print-tip effects, Tm,position in gene, etc. We would hope that biological changes would be independent of technical characteristics of ex