Why MA plot should be normalized?

[tedwong]

I'm talking about the slide: http://bioinformatics.mdanderson.org...09/ma08_bw.pdf

In page 5, the title is "Why is Normalization an Issue?". The author compares the MA plot before and after normalization. He writes "Variation that is obscuring as opposed to interesting."

Q1: I can see the fitted red line before and after normalization. What's that red line?

Q2: Why is the second MA plot (after normalization) is better? The distribution is more symmetric, but what does that have anything to do with biological variation and statistical testing? Why should I care symmetry?

[dpryan]

1) The expected value of the data points as a function of A.
2) We don't expect there to be a global unidirectional shift in changes, rather some genes should go up and others down.