Aim and experimental design

Since the GeneChip probe set sequences are strongly biased towards the 3'-end of the published sequences, it is implicated that under-representation of 5' mRNA sequences occurring in cRNA targets from moderately degraded RNA samples may not affect data quality in most experiments. However, a systematical experimental analysis to what extent this might create false positive targets is lacking. To determine the impact of degradation on microarray data, aliquots of an intact RNA sample prepared from HeLa cells were treated in a controlled fashion by heat in the presence of divalent ions, which results in random cleavage of RNA molecules. After chilling, equal amounts of a set of intact poly(A)+ RNAs were spiked into partially degraded as well as untreated control samples. These spiked-in RNAs, which can be measured on the arrays, served to monitor various steps of the enzymatic conversion into cRNA and, importantly, as external normalization controls using the selected probe set scaling option. All targets were prepared and analyzed by hybridization to Affymetrix HG-U133A arrays containing more than 22 000 probe sets. Since the overall sequence content in degraded and control RNA samples remained identical, genes that were identified by statistical analysis to be significantly up- or down-regulated in the degraded samples represent false positive targets, which are solely due to the introduced 3'-bias in the cRNA target from partially cleaved RNA. In addition, we compared the effect of two different array normalization procedures, global scaling and selected probe set scaling to spiked-in controls, on the rate of false up- and down-regulated transcripts.

0 0

Post a comment