What is the best normalization for microarray data?
I have microarray data from the Yeast 2 chip from Affimetrix. I know I need to normalize it. But what kind of normalization should I use? Some people recommended RMA and other quantile-quantile normalization. My results are changing based on what kind of normalization I use. But the analytical methods should not impact the results? Or should they? How can I determine which is the best kind of normalization to run and how can I justify my choice in the method section of a submitted publication and in the defense?
Does Affymetrix has its own analysis and normalization software? If so, where can I download it?
There is no such thing as 'best' normalization. There are different methods, and each has its own underlying assumptions. Which you use depends on what you are willing to assume, given the data in hand.
In addition, please note that RMA isn't just doing a normalization. The algorithm includes a background correction, quantile normalization, and then a summarization step. I would imagine this is the most common summarization method (and you can use RMA with Affy's software as well). As for defending your choice, that was probably more of an issue maybe 15 years ago when people were really trying to figure out what methods seemed to do the most reasonable job. These days the people I work with just say they used RMA, cite the original paper and get on with it, and I have yet to see a reviewer balk at the choice.
And Affymetrix does have its own software, which you can find at affymetrix.com.
