What is the difference between deferentially expressed genes and deferentially expressed transcripts?

Question

[WUSCHEL]

I know the Biology behind deferentially expressed transcripts.

But I do not understand what does it mean / difference between deferentially expressed genes and deferentially expressed transcripts / isoform levels?

In my RNA-Seq analysis, there are sig. difference data at DEG levels, but those genes are not present at the transcript level. Is this possible? or my analysis went wrong?

Hypothetical e.g.:

@ At1g16610 ; I have sig results

But @ At1g16610.1 /At1g16610.2 / At1g16610.3 etc nothing is significant.

** I am a wet-lab person, learning bioinformatics thanks to this great community at BIOSTAR

[Pyretu]

For better understanding: Which tool did you use for your analysis? What do you mean in your example? Do you mean that on transcript level the single ones are not significant but in a combined "whole" single "thing" they are significant?

[WUSCHEL]

Yes, on transcript level the single ones are not significant but in a combined "whole" single "thing" they are significant? I.e. at Gene level , it is significant, but not at isoform level.

This is my analysis pipeline

[Pyretu]

Thank you for the clarification and the overview of your analysis pipeline. Sadly this problem exists and depends on the tools you use. There are tools that just "split" the fpkm values between the isoforms such that every single one has the same, some others are calculating the a little bit weirdly. The problem regarding isoforms is, that you usually do not know, from which isoform the exon origins (for exons used in 1,2,3... all isoforms) so you have to make assumptions on that. I really do not know on the fly what your tools are doing on this topic because I personally use completely different. What you could possible do is, have look at the The Integrative Genomics Viewer (IGV) to have a look on the read distribution.

[dsull]

Not sure I understand this response. If doing differential expression, you shouldn't be using "fpkm values" in the first place (so unless you're using some outdated tool, there shouldn't be any "splitting" of "fpkm values" evenly between isoforms). Most tools nowadays should be able to reliably estimate which isoform a read comes from.

[dsull]

https://liorpachter.wordpress.com/2018/02/15/gde%C2%B2-dge%C2%B2-dtu%C2%B2-dte%E2%82%81%C2%B2-dte%E2%82%82%C2%B2/

In answer to your question: Yes, it is possible. Per the source above: "A key issue with DTE is that there are many more transcripts than genes, so that rejecting DTE null hypotheses is harder than rejecting DGE null hypotheses."

[WUSCHEL]

Thank you dsull. Appreciate.

[WUSCHEL]

Thank you Pyretu. Appreciate.

Answer 1

[Kristoffer Vitting-Seerup]

There are 3 ways changes in expression can occure in a gene with multiple transcripts:

1. All expressed transcripts can be regulated in the same direction (up or down) giving rise to a differential gene expression. In this case the individual transcripts do not necessarily have to be differentially expressed by themself for the gene to be differentially expressed (since the many smaller individual transcript changes when added together becomes a larger change) - but naturally they can be.
2. A subset of expressed transcripts can be regulated. An easy example is that a single transcript is upregulated while the rest are unchanged aka a differentially expressed transcript. This could also lead to a differentially expressed gene if the transcript is one of the most expressed in that gene or the change is very large. But if the transcript is lowly expressed (compared to the other transcripts) this might not cause the gene to be differentially expressed..
3. The third possibility is an isoform switch where one transcript is upregulated while another transcript is downregulated. In this case the transcripts might be differentially expressed - but since they cancel each other out on a gene level the gene can remain unchanged.

It is however important to remember that these events are not mutually exclusive. You can easily have an isoform switch in a gene which is also differentially expressed.

Hope this helped.

Cheers Kristoffer

P.s. if you are interested in isoform switches my R package IsoformSwitchAnalyzeR is made to identify isoform switches with predicted functional consequences both in individual genes and on a genome wide level - and it works directly from Salmon/Kallisto data as well. Via these links you can see examples of the gene level analysis and genome wide level analysis.

[WUSCHEL]

Thank you kristoffer. Appreciate.

[Arindam Ghosh]

Is there any specific example for the point 3. I was searching for a gene that could be DE at transcript level but not a the gene level.