Hi,

I did RNAseq (single end) for a drug treatment (Drug A) on a cell line to look for the gene expression changes. From the literature, i found some Microarray data on the same cell line treated with other drugs (Drug B, C, D).

I analysed the data of microarray data of this Drug B,C, D samples separately using the limma package and got the differential expressed (DE) gene set Similarly i analyzed RNAseq data using Deseq2 to get the DE geneset. i would like to compare this DE genesets from different treatments to identify the common and unique genes for every treatment.

Here are the questions/doubts that raise in my mind,

1. Can we make the comparison between microarray DE and RNAseq DE ? (having aware that microarrays have defined probe set correspond to only few genes)

2. If we can compare, do they have to be analyzed in a similar way?

3. If they can be analysed similar way, how to make the cross platform comparison to show statistically that these DE which are common in this two or more treatments and these genes are specific for only this treatment?

I have gone through some threads of discussion in the forum but have not answered my questions.

Please feel free to ridicule and suggest me

Thanks in advance

1. Yes. But it depends on what you mean by 'make the comparison'. You could certainly look for genes that are differentially expressed in either comparison, using say Venn diagrams. But that is sort of boring and naive. If you get ten genes that are consistent, what does that mean, exactly? There are other things you could do, for example looking for gene sets that are consistently affected in two or more of the drug treatments, particularly if you have an a priori hypothesis as to what pathways might be affected by all the drugs. A simpler thing to do would be to say the differentially expressed genes for drug A are the gene set you care about and then do self-contained gene set tests for those genes in the other three comparisons.
2. Another alternative would be to use either the geneMeta or mogsa packages (or possibly others that I don't know about), where you combine results. This would require you to make an assumption about how similar the four drugs might be.
3. See the vignettes for geneMeta, mogsa, and the limma User's Guide (plus the help for roast and fry in the limma package).