Please have a look at the paper from SYLVAIN PRADERVAND and colleagues (RNA (2009), 15:493–501) who have investigated different normalization algorithms on the Agilent plattform.

We are working with the Affymetrix system where Quantile Normalization gives quite comparable results with respect to validation by qRT-PCR.

Some months ago, I have done normalisation using RMA method in. You can see the comparison of it with other methods.

As in RMA the quantile normalization step is carried out at the probe-level, rather than the probeset level in other method. This means that the distribution of probe intensities is identical across arrays before the median polish summarization.

Thank you for your answers! Have read Pradervand et al. adn found it very useful. Christian: I have used the AgiMicroRna package. The handbook suggests normalization before filtering of controls and non-epressed miRNAs. After filtering, the density plots and other QC plots look very different: it no longer looks properly normalized. Is this OK, and to be expected given that i'm left with about 150 expressed miRNAs? Or should normalization be carried out after filtering to avoid too much noise from poorly expressed miRNAs?