How can a short read be assigned to a particular transcript isoform?

[AlanRaetz]

Okay, first post, I hope I can explain this. I am interested in transcripts that encode ERCC5 isoforms. and according to this (and other sources), there is the full-length isoform and a number of short isoforms.

http://www.ebi.ac.uk/Tools/geuvadis-...1589-103543957

I'm just puzzled by the basic conceptual problem that if all that you get out of rna-seq data is short reads, how could you possibly assign a short read to a particular overlapping isoform? For example here, there is the full-length isoform, and also short 5' and 3' isoforms.

Does this rely on polymorphisms within individuals?

Or is this simply based on the quantity (count) of reads in certain regions, ie, if you have 2x reads at the 5' end and also 2x reads at the 3' end, but a region of 1x reads in the middle, then you fit this data to a model where there is a short 5' isoform and a short 3' isoform and that would explain the lack of reads in the middle of the gene.

Hope the question makes sense. Let me know if I can clarify.

Thanks! -Alan

[dpryan]

It's a pretty non-trivial thing to do and I don't think we (as a field) have arrived at a perfect answer yet. Your best bet would be to read through the cufflinks and rsem papers to see how they approach the problem. Depending on what you really want to do (e.g. if you're interested in differential isoform expression due to some treatment, or between tissues), you might find the "differential exon usage" route easier to understand. In that case, have a look at DEXSeq and the paper for it. That will often yield useful results in a manner that's easier to follow (and therefore debug and check for coherency...always good things to do).

[AlanRaetz]

Thanks

Devon, thanks for the pointers as far as looking into this some more.

I was a C/perl software engineer in a former life, now I have to code for fun and do wet lab work in my postdoc.

These 5' and 3' isoforms that are annotated as protein coding are very interesting to us, as the protein folds into a form where the middle region is a linker and the catalytic domains form from the N- and C- terminal ends.

Rare genetic loss of XPG leads to premature aging and XP; the interesting thing is that expressing the full-length protein in cells that lack XPG does not fully restore wild-type activity of endogenous XPG. So maybe these short isoforms encode proteins that have some additional or different functionality.

My plan is to verify these short cDNA isoforms exist in human cells by RT-PCR first. If they do, I'd try to express the short cDNA from plasmid constructs and see if the proteins are stable.