Figure 7

Effect of non-dominant invariant genes on data normalization. (A) MA plot of baseline array versus Dnl data containing 30% up-regulated genes. Each blue point represents a feature in a microarray. Red and black dots mark the expected values of the 1^st and 2^nd largest peaks of LER density distribution for each interval by NVSA method, respectively. Green labels each NVSA-fitted normalization value. Black arrow points to the center interval (red circle) of the seed invariant in NVSA analysis. (B) Fitted normalization values by cross-correlation method on the same microarray data (Green). (C) Number of non-dominant invariant gene classes that occur under each gene effect condition of Dnl data set. Invariant gene class is defined as non-dominant when the class is not the largest peak of LER distribution in NVSA analysis. It is noteworthy that the two or three non-dominant invariant gene classes shown in data containing 2.5 – 50% down- or 2.5 – 40% up/down- regulated genes are all located in the first two or three intensity intervals where normalization errors are not 100% included in the calculation due to the exclusion of boundary data points. Thus in these data, the effect of non-dominant invariant genes may not be reflected in normalization errors.