Comparison between two indvidual tissue samples from RNA seq

[shikha5]

Hi Everybody
I have a question regarding the RNA seq data analysis in terms of the differential exon usage by two different tissue samples such as comparison between kidney tissue (RNA seq reads aligned to reference genome) and liver tissue. I want to look for the differential exon usage like if some gene xyz in kidney tissue using say exon 1 but same exon is absent in liver gene xyz.

I welcome all your valuable suggestions related to my query.

Thank you.
Have a good day.

[dpryan]

If you really only have 2 samples (and not, as in your example, say 3 liver samples and 3 kidney samples), then there's no reliable way to do that. You can certainly find differences between the samples, but you can't then know if those differences are representative of tissue differences or not. If you have replicates, just use DEXSeq.

[shikha5]

Thanks dear for your reply and suggestion.

Okay, I also think that this is a good idea to consider the replicates of a sample but if I use the DEXSeq tool then it will only tell me about the variability among replicates of a same sample not between two individual samples with replicates. For example two samples with 3 replicates each, now the question is: Can it tell me the variability in terms of their "exon usage" between two samples considering all their replicates?

[dpryan]

Yes, DEXSeq considers all biological replicates of a group in the comparison.

[Jeremy]

It sounds like you have paired samples (two tissue types from a single individual), is that correct?
I haven't used it, but I think EdgeR supports paired samples.

[dpryan]

@Jeremy, edgeR is intended for differential expression whereas DEXSeq is intended for looking at differential exon usage. You could use edgeR (or even DESeq) for this sort of thing, but it would take more effort than needed. DEXSeq uses a generalized linear model (like edgeR/DESeq/etc.), so it can also deal with paired samples.

[Jeremy]

The result of a differential expression analysis depends on what the input is, if it is gene level information you get differential expression results of genes. If it is exon level information you get differential expression results of exons.

[dpryan]

The result of a differential expression analysis depends on what the input is, if it is gene level information you get differential expression results of genes. If it is exon level information you get differential expression results of exons.

Yes and no. If you directly used, say, edgeR without any modifications, you would have many more false positives, since most/all of the exons of a differentially expressed gene would show significant differential usage, even though this would be an artifact. One could simply write code to handle these situations in edgeR, but then you're mostly just recreating the wheel/DEXSeq.

[Jeremy]

Well, my point was that if he has paired samples then the analysis needs to be for paired samples, edgeR was just an example.