Reliability of mRNA profiling: verification for samples with different complexities.

Normalization of mRNA profiling data remains an open issue, which turns critical when comparing divergent samples or mRNA populations with different complexities. To address this question, we generated samples with different RNA amounts and complexities by subcellular fractionation of cytoplasmic RNA into the mutually exclusive ribosome-free and polysome-bound RNA pools. For each of the 563 mRNAs analyzed, the hybridization signal corresponding to the cytoplasmic sample equals the sum of signals from the ribosome-free plus the polysome-bound targets (cytoplasmic mRNA = ribosome-free mRNA + polysome-bound mRNA). This intuitive equation was fulfilled only after data normalization following "spiking" of the samples with an exogenous RNA. This is the first demonstration that spiking allows one to correct not only for differences in reaction efficiencies but also to reflect the variations in amount and complexity between the initial mRNA populations.