Upstream Regulator Gene Differential expression

Question

[benjyrolls]

Hi

I performed an RNA Seq on two conditions SOX2 WT mouse and SOX2 KO mouse to find the genes which were differentiallly expressed through an RNA Seq pipeline and eventually obtained the differentially expresses DE genes by Cuffdiff

To find the upstream regulators these differentially observed genes were uploaded into both IPA and CHEA2 However I found other transcription factors as the main players in these differentially expressed genes as compared to my transcription factor SoX2 .

My question is does it mean My Sox2 experiment failed and why are my seeing other transcription factors as the main players in DE genes other than SOX2?

Thanks

Benjy

Answer 1

[Devon Ryan]

I'm not sure why you started looking for other transcription factors when you explicitly knocked down SOX2. If there are known/presumed SOX2 binding sites in the promoters of most/all of the highly DE genes then it doesn't matter what some random tool tells you, you already know that the SOX2 KO is driving the response.

The things with all of these tools, is that they rely on (A) a database for known target genes (how updated is this? who knows?) and (B) the test is looking for enrichment. The target genes in the database will end up being "all known target genes", which of course will not all be expressed to begin with in your circumstances. So it could well be the case that if these tools knew of the subset of target genes expressed in your particular circumstances that they would give you results more in line with your expectations.

Anyway, when using a tool it's always good to ask yourself, "is this really answering the question I want to pose, or am I only getting the answer to a related question?"

[benjyrolls]

Well am very new to RNA Seq because I have a wild type mouse with Sox2 injected(condition 1) and other with SoX2 removed(condition 2) I was thinking the differentially expressed gene will be as a result of Sox2?

Can I say that Sox2 presence is driving the response and the DE genes are as a result of SOX2 presence?

[GouthamAtla]

How many replicates do you have ? What does the PCA tells ? Does it clearly separates the WT and KO ? Does the GO of these differential genes agree with the biological function of SOX2 ? Instead of looking for IPA etc, you can do a motif enrichment analysis ( Homer/FIMO etc ) on the promoters/proximal promoters.

As Devon said, the genes that are regulated by SOX2 really depends on the particular experimental conditions, so you can not completely rely on the databases. I would do a motif enrichment analysis on the promoters of differential genes.

[Devon Ryan]

As long as you don't say "directly" then sure. If you want to additionally make the argument that the expression changes are directly due to SOX2 (as opposed to SOX2 changing the expression of something else that in turn drives much of the expression differences you're seeing) then you need to confirm that it can modulate the expression of those genes. Checking for known binding sites in the majority of the DE genes should reasonably suffice for this.

[benjyrolls]

There are 3 replicates.

Well, I was trying to reproduce the results in this experiment using the RNA-Seq data however am not confident in the differentially expressed genes and its role to Sox2 in terms of RNA-Seq.

However in the ChIp-Seq experiment, Centrimo motif analysis, using the Chip-Seq data shows Sox2 as the master transcription factor regulator.

[benjyrolls]

As per you statement "Checking for known binding sites in the majority of the DE genes should reasonably suffice for this."

I have performed a check on the binding sites and this is what I found Sox2 seems to be regulating some genes in conjunction with others?

[Devon Ryan]

If the ChIPseq experiment agrees with the RNAseq experiment then you don't need to bother with any of this. That already shows that the change is due directly to the SOX2 manipulation. Sure, someone could argue that maybe it's really indirect in some cases, but the only way to really test that is to remove the binding site in each case and see if SOX2 modulation still has an effect (or maybe add in an inactive SOX2, if such a thing exists).