workflow for combining multiple microarray studies of same platform

Question

[CHANG]

I am pooling together tumor samples of of 4 different histology from 4 GEO studies that are done on Affymetrix U133 Plus 2.0. I want to compare differential genes and between each tumor histology and blood.

Sample Tally

study   tumor_histo_1   tumor_histo_2   tumor_histo_3   tumor_histo_3   normal blood
GSE1    5               5                                               10 
GSE2    5               5                                               5
GSE3                                    10                               
GSE4                                                    10              5

Here are my proposed approaches

1. Use insilicoDB to get fRMA normalized expression sets for each study so that all study will have been normalized the same way (Description: http://bit.ly/1OMmRJx). Use COMBAT to remove batch effects. Then run limma to get DE genes for each tumor histology against the pooled blood data.
2. Pool together raw cell files from all studies, run RMA, then limma to get DE genes for each tumor histology against pooled blood.

Thank you very much!

Answer 1

[Sean Davis]

fRMA probably offers some advantages over straight RMA, particularly if you are planning to add more samples/studies down the road. That said, the batch effects are confounded with your variable of interest (tumor histology), so your results will just need to be interpreted carefully.

[CHANG]

Thank you in advance.

[bcbio_uk]

Hi Chang

I am keen to combine multiple GEO datasets (all run on Affymetrix U133 plus 2.0) and came across your thread. I was wondering what approach you ended up using in order to combine your datasets? I would appreciate any help.

Thanks

[CHANG]

What about pooling all the raw cell files from studies and run fRMA, is there any advantage of doing that over fRMA on individual studies then run COMBAT?