RNA amplification, fidelity and reproducibility of expression profiling

We quantitatively address the effect of T7 RNA amplification on expression profiling data, and answer the following questions: (1) What fraction of genes sampled is amplified non-linearly? (2) What is the effect of RNA amplification on comparative expression measurements? (3) If there is amplification bias, is the bias dependent on the degree of amplification or the amount of starting material and (4) Does amplification increase the overall variability of the results? We show that while there is significant amplification bias, the bias is consistent and generally has little effect on array comparisons between amplified samples.