Annotate RNA-seq transcripts to transcript class as NMD (nonsense mediated decay)

Question

[Chirag Nepal]

Hi there,

I would like to annotate if assembled RNA-seq transcripts/isoforms represents nonsense mediated decay transcripts or not. Any suggestion on how to do this? Or are there any existing tools to do so?

Thanks

Answer 1

[Dave Tang]

Hi Chirag,

my approach would be to use Gene Ontology. For example, I found this GO term (GO:0000184) and its full name is "nuclear-transcribed mRNA catabolic process, nonsense-mediated decay". I would then proceed to find the genes that are associated with this GO term, and then find the transcripts associated with those genes. Those transcripts would thus be associated to NMD.

You can search through all the other GO terms to see if there are others associated with NMD.

Cheers,
Dave

[Pierre Lindenbaum]

I think he wants to find the genes affected by NMD, not the genes involved in the NMD mechanism..

[Chirag Nepal]

I want a systematic way to predict, if novel RNA-seq transcripts (or lncrna) are NMD transcripts or not.

So far, my approach has been to look for stop signals (in the exons of transcripts). But the problem is, as start codon is unknown, it might be difficult to tell if the encountered stop codon is in the right ORF.

[mfahim]

Hi Pierre/Chirag,

I am in a similar situation. Can someone let me know what parameters should I use for cufflinks and the subsequent analysis. I have a biology background and am using Galaxy for my analysis, with very limited knowledge of command prompts.

@Chirag,

I would appreciate if you can update your success story.

[Chirag Nepal]

In my opinion, simply with only RNA-seq, it is almost impossible to predict if the transcript qualifies for NMD.

Answer 2

[A. Domingues]

As far as I know there are no existing tools to do that 'off the shelf'. What you do however, is to take datasets where targets for NMD have been detected and intersect with your data. There a few knock-down experiments of components of the NMD pathway,, and also ClIP data. For instance (quick and dirty google search):

• http://compbio.berkeley.edu/people/weig/RNA2011-UPF1-RNASeq-Poster.pdf
• http://genome.cshlp.org/content/early/2013/06/13/gr.157354.113

You might have to re-analyse these datasets, i.e.., performing de-novo tranbscriptome assembly in the UTF-1 ko dataset, and finding which transcripts are more expressed in the absence of UPF1. These you can then use to intersect with your data.

This is of course not ideal, and you will have to be lucky enough to be working with the 'right' species, but I think it would go a long way in predicting transcripts as NMD targets or not. On the other hand, it will be based on biological observations, so for my money a more reliable than simply looking for PTC. When I was doing a lot of cloning, I found that some of these PTC-containing transcripts escape NMD and produce functional proteins,

[Chirag Nepal]

Thanks for the papers. I am definitely gonna re-analyze those data sets. The reason why I wanted to check PTC, was to test, if subset of novel transcripts (we believe these are NMD transcripts, based on exp data) are enriched for PTC, compared to other set of transcripts. Alternatively, I can intersect with transcripts from analyzing the data from the paper you send.

BTW, you said some NMD transcripts produce functional peptide. Interesting ! But I assume these NMD transcripts had overlapping isoforms, which was meant to produce functional peptide. Perhaps, if that is the case, the major isoforms might influence the NMD transcripts?

[A. Domingues]

I mentioned 'PTC-containing, and not 'NMD transcripts', subtle but important difference. Regardless I apologize because this statement is not correct:

When I was doing a lot of cloning, I found that some of these PTC-containing transcripts escape NMD and produce functional proteins

I went back to my data, and the PTC-containing transcripts that I cloned and tried to express did not produce a functional protein. There some resulting from odd splicing patterns, 5' and 3' ASS, that were funtional.

Btw, there might other datasets out there, those were just the ones that came up in a quick search.

Good luck. I find 'aberrant' splicing and NMD fascinating.