Majority of genes upregulated in RNA-seq study of toxic compound

Question

[mschmid]

I am currently doing an RNA-seq study that deals with the effects of a toxic compound.

It is not very surprising to me, but most genes are up regulated when the organism (insect) is exposed to the toxic compound. What can people say having more experience with this? Is this the general picture.

I am looking for a review/paper that deals with this to read more. Do you have a good suggestion?

EDIT, some more information:

We study the effects of a toxic compound on insects.

We have Illumina-sequenced the Transcriptome of: 5 biol. Replicates for Control, 5 biol. Replicates for 1x concentration of the compound and 5 biol. Replicates for 10x concentration of the toxic compound.

First I did a Trinity-assembly using all 15 biol. replicates. Then I map the the data back to the assembly using bowtie2 and then use RSEM. After that I use Trinity scripts to calculate DE (using EdgeR) and to output the significantly DE genes.

For both conditions, I see mostly up regulated genes. For the 1x concentration about 60 (50 up regulated), for the 10x concentration about 280 (230 up regulated). This makes sense, the number of of diff. regulated genes correlates with the concentration of the toxic substance.

If we check the overlap between the DE genes of the two treatments, I have about 25 DE genes. Those are ALL up regulated.

So, would you say that this compound mainly up regulates genes or could this be an artefact of the de novo-based method?

[Benn]

Are you assaying all genes? Or do you just have a subset of genes? I mean do you have a reference genome? Or is this based on de novo assembly?

[mschmid]

It is based on a de novo assembly. Sorry, important information :)

[Benn]

When you look at de novo assembled data, the dataset is biased. You only have genes present in both situations (the treated and untreated samples). So it is not a surprise to me either.

[mschmid]

I am not sure if I can follow you. True, the data is biased...

But the de novo assembly is based on all conditions. So if there is a down-regulation of a gene in the treatment, I should still see down-regulation. But in my case the vast majority (about 80%) of DE genes is up-regulated in the treatment. This looks more like a biological thing, no?

[Benn]

You should explain more about your approach, it is difficult guessing what you did...

[mschmid]

OK, added some more info. Hope it gets clear now.

Answer 1

[Benn]

Thanks for the edit, like I said in my comments, your dataset of transcripts is biased. They are enriched for your toxic compounds. Genes that are down regulated due to your compound are not expressed and therefore not assembled (only a few transcripts that have enough copies in the untreated control samples will be assembled). You might also lose downregulated genes by a filter which is often used in edgeR (don't know if you did filter?). That you see more up than down is caused by the biased dataset, if you would have the whole genome it could be different (more down than up, hard to predict).

[mschmid]

Hmm... OK. Like this it makes sense. That would mean Ideally one has a bit more Replicates for the control to counter for this effects?

[GenoMax]

You should clarify how you treated your assembled transcripts. Did you filter them aggressively to get the number down? You may have removed some low expressed genes by doing that.

[Benn]

Or sequence deeper to get more transcripts only present in control.

[mschmid]

Yes, sure, sequencing deeper would be the easier method