Figure 12

Tissue-specific RNA profiling. Part a) Frequency distribution of absent calls of tissue-specific total RNA hybridized on HG-U133A arrays taken from 79 tissues and analyzed with MAS5 and hook (raw array-intensities and MAS5 data were taken from ref. [44]). Part b) – g) Comparison of two hybridizations with small and large absent rates (see arrows in part a): peripheral blood-BDCA4 dentritic cells (dc, GEO-query GSM18873) and superior cervical ganglion (scg, GEO-query GSM19012). Both hybridizations used the same amount of total RNA (5 μg) for synthesis of biotinylated cRNA and the same labelling protocol. Part b) compares the log-intensity-distributions of the PM-probes: Except the shift and widening of the distribution of the dc-sample, one observes essentially no peculiar differences between the specimens. The median and the probe-set related values of BioB-3 are explicitly shown and discussed in the text. Parts c) and d) show the respective hook-plots together with the signal-density distributions. Note the striking differences: The scg-sample hybridizes much weaker with a markedly larger fraction of probe set with absent calls (95%) and a much steeper decay of the distribution in the mix-range of the hook (λ is the decay constant). The dotted curves are fits of the Langmuir model. The open symbols indicate the hook-coordinates of selected probe set in both preparations to illustrate the apparent expression changes from different regions of the hook (#1 to #4). The solid symbols refer to amplification and hybridization control probe sets. In part e) and f) the Σ-coordinates and the expression measures of these selected probes are explicitly shown, part g shows the respective log-differences between both samples. Note that the difference of the non-specific background level is negative (dashed horizontal line), whereas the difference of the specific binding strengths of most of the considered probe sets is positive (PMonly measures). The specific expression of the BioB-control is virtually invariant in both samples, as expected. Contrarily, MAS5 pretends significant expression changes of the BioB-control due to improper normalization (see text).