Affymetrix Power Tools To Measure Whether The Gene Is Expressed Or Not?

Question

[GR]

Dear All,

I have few genes that I want to know are expressed or not in few tissues and conditions (microarray data). I was wondering if anyone has used affymetrix power tools to detect the signal intensities to say whether a gene is expressed in the given condition/tissue etc. I got the normalized signal intensities but I don't know what threshold to use to say a particular gene is expressed or not.

Any other way/steps to do this will also be helpful.

Thanks, RT

Answer 1

[Ido Tamir]

• mas5:

In mas5 there is the stratification into a (absent), m (marginally), p(present). I think the mas5 normalization is not very good (underestimates the low range (or was it overestimates ?, I forgot)). The a,m,p classification however is rather accurate.

• GCRMA:

You could normalize with GCRMA, then plot the density of expression. You normally see a huge peak that represents the not expressed genes, then expressed genes come up. 2 distributions: low end = background intensities of not expressed genes which fluctuates higher = intensities from expressed genes. Choose a cutoff somewhere where you think most of the data comes from the expressed genes. In the data shown I would say that between intensities 4-5 probes (not genes!!) might be detected as expressed (~ marginally in mas5) and above 5 they are reliably expressed and detected.

[GR]

Hi Ido,

Do you know if there is any way to extract p-values for the present/absence calls. I could just get the P/A/M labels using the following R code. I want to add p-values to these results. Any suggestions?

Thanks, Reetu

affy.data <- ReadAffy()

data.mas5calls = mas5calls(affy.data)

data.mas5calls.calls = exprs(data.mas5calls)