RNA quality, while the RNA quality from nondiabetic samples is
acceptable. It is possible that there is contamination during sample
collection for diabetic samples. Although these arrays are not per-
fect in quality, they can be used for the subsequent analysis.3.2 Normalization The noisy signals from the experiments should be removed by
data normalization. The intensities from arrays are processed
through background correction, normalization, and probe-specific
correction. The probe set-level data are on log scale. There are
several methods of background correction, normalization, probe-
specific correction, and summarization provided byaffypackage
[32]. The background data can be corrected byrma method
(Fig.6) but should be conjugated withpmonlymethod for probe-
specific correction. The data summarized bymedianpolishmethod
should not be used aftersubtractmmmethod for probe-specific
Fig. 3Intensities of arrays. (a) R code for showing the intensity of all arrays. (b) The intensities of all arrays are
shown in box plots
Building Biological Networks from Gene Expression Data 183