Duplicate genes with same gene symbol annotation

[Jason Toy]

Hi Brian,

I ran into a bit of a theoretical hiccup in my analysis and am hoping you can guide me on the best practice here.

I am currently trying to run enrichment analysis with GOseq and I am realizing that my annotated expression matrix (gene level) has a lot of duplicate genes (I annotated it through Trinotate using blastx against swissprot). Most of the time these duplicates are all sequential too. E.g. gene IDs XLOC_000044 - XLOC_000049 all annotate as gene symbol ATRN1 (Attractin Like 1). The transcriptome assembly was done with a reference through tophat/cufflinks.

This presents an immediate problem because GOseq won't allow duplicate gene names, but it also poses a bigger question of "should these all be grouped together since this is a gene-level analysis?". Looking at a number of duplicate cases, they tend to show similar DE patterns across duplicate genes. In other words, should I merge the rows for each of these into one for each gene symbol? And if so, at what point during the analysis (raw counts? After TMM normalization?)

Thanks in advance!

Jason

--

Jason A. Toy | PhD Candidate

Rent Burden: 41% (what’s this?)

Dept. of Ecology and Evolutionary Biology

University of California, Santa Cruz

Global Change Biology Lab

Molecular Ecology Lab

Office: (831) 502-8031

    

[Brian Haas]

Hi Jason,

If all these 'genes' are actually aligning at the same genomic locus, then they were probably just not annotated properly as isoforms of the same gene, and you'd need to regroup them.  This could easily be a cufflinks issue.

best,

~b

--
You received this message because you are subscribed to the Google Groups "trinityrnaseq-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to trinityrnaseq-u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/trinityrnaseq-users/CAKaDAZzh7oUug7njN7dW7pA%2BT762yU8OdQVAxxvU6goT5rti9Q%40mail.gmail.com.

--

--
Brian J. Haas
The Broad Institute
http://broadinstitute.org/~bhaas

 

[Jason Toy]

Hm, okay. How would I go about confirming if these 'genes' are actually aligning at the same genomic locus?

And if I need to regroup them, should I do this by summing their rows in the raw counts matrix (RSEM.gene.counts.matrix) or after running edgeR's calcnormfactors()? Or does it not matter?

Thanks again,

Jason

[Brian Haas]

yeah, if it's cufflinks, then you'll have genome locations for everything.

You'd sum them up to the same gene if you're working with RSEM's estimated counts (not raw counts).  Then, you could redo your edgeR analysis.

[Jason Toy]

Okay, gotcha. Is the RSEM estimated counts the RSEM.gene.counts.matrix? Or a different one?

Thanks so much!

[Brian Haas]

Given the redefined trans/gene mapping file, you'd rerun the abundance estimation matrix building step, and that'll give you the refined gene matrix.

The isoform matrix has the estimated counts.  The gene matrix involves some additional math that involves some weights - details on the wiki.

best,

~b

[Jason Toy]

Hi Brian,

After manually investigating my annotated transcriptome, it seems that most, if not all of these cases of duplicate annotations are due to split genes. Each of these putative genes map to a different non-overlapping portion of the annotation reference gene. So these appear to be cases where many putative genes are actually one single gene. Is there a good tool to automatically detect and merge split genes across my entire transcriptome?

Thanks!

Jason

[Brian Haas]

Hi Jason,

PASA will merge annotations if there's transcriptome evidence for them to be merged, given a reference gene annotation and the annotation update functionality.  This may not be what you're after, though, and could be overly complicated to address this, plus I'm not sure how comprehensive it would be - solving some but not all cases.

You might need to roll your own solution here.

[Jason Toy]

Alright, ya PASA might not be what I'm looking for on this. Thanks for your insight Brian!