Meta analysis of expression microarray data

Question

[hntiger1688]

Dear all,

I am doing a project about meta-analysis of GEO microarray data. But I encountered a problem:

In the microarray data, there are more than one probes to detect the same gene and therefore there would be several expression data of the same gene. But how can I integrate it? I used to get the mean of the data but I found that it was not good because some data with bad quality will interference other data.

So can you tell me the correct way to integrate the data?

Thanks

[Neilfws]

why is this tagged chip-seq if the question is about microarrays?

Answer 1

[Jason]

You are probably dealing with affy arrays. Here is a tutorial on how to use R in order to analyze such data.

Just to clear up any potential confusion, although I provided a link that is on the HOMER website, you won't use HOMER to solve your problem. HOMER is a suite used for seq data. I guess they just thought it was useful to include a workflow to analyze microarray data on their site too.

p.s. This is just a general way of analyzing expression data. You should probably refer to the methods section of the papers you are getting the data from to see exactly how the authors performed the experiment.

Answer 2

[Manvendra Singh]

Once, you have processed your data and have got expression value for the probes significantly. Afterwards, I would define gene expression value by assigning mean of expression values of probes corresponding to that gene. In this way, you would have one expression value for one gene. Consequently, I would merge the dataframes by their gene names. If they are different platforms, then remove the batch effect by quantile normalization. The dataframe would be ready for your analysis.

[hntiger1688]

Thank you! But how about P-value and logFC. So you mean that I need to integrate the expression data first and calculate the P-value and logFC. Is that it?

[Manvendra Singh]

You could calculate Z-scores across the rows