Differential Gene Expression between 2 RNA seq groups without normal control

Question

[bionerd]

Hi we performed RNA seq on 8 mammary tumors. I wanted to compare our RNA-seq data to raw RNA-seq data that I found online. I plan to process the raw RNA-seq data through the same pipeline. Is it possible to run EdgeR, Deseq, Limma or any other programs to find differentially expressed genes between the two datasets? My main concern is getting rid of batch effects. The issue is that I don't have any normal mammary samples on my RNA-seq to use as control. There is normal and tumor samples on the RNA seq dataset that I downloaded. Anybody have any suggestions?

Thank you!

[bionerd]

Thanks for responding. Initially we wanted to generate RNA-seq data so that we can compare it to microarray data that we already had. But unfortunately we realized comparing RNAseq and microarray data is very difficult. So we are trying to compare it to a RNAseq dataset we found online. We know the experiment design is flawed but we really don't have the resources to repeat the experiment right now so we are hoping to get as much information as we can from our current data. Any suggestions would be appreciated!

[russhh]

Is there any substructure here? Are they 8 randomly selected mammary tumours, or are there 4 of one subtype and 4 of another (or something similar).

Answer 1

[andrew.j.skelton73]

Finding data online to use as a control will not produce favourable results, even if by chance you get the correct type of samples you're looking for. You're correct to be concerned about batch effects as you won't be able to tell what is biological variance, vs what is technical variance (instrument, prep, temperature of the lab, time of day it was extracted, alignment of Jupiter, etc).

What I'd usually suggest in these cases is non parametric comparisons, such as rank prod where you can compare two differential expression results, however if you have no controls on your RNA seq experiment then that's out of the question.

What was the purpose of your experimental design?