Repository navigation
Isoform nomenclature question #217
Description
Hi EGAPx team! I have some questions about the isoform nomenclature rules that EGAPx uses in product attributes. I've noticed three types of isoform product suffixes (where 'N' is a number):
- isoform N (for mRNA/CDS features)
- -TN (for mRNA/CDS features)
- transcript variant XN (for transcript features)
Some observations about the suffixes:
- In some cases, the number used in the 'isoform' suffix and the '-T' suffix is different for the same isoform.
- For protein-coding isoforms, the transcript ID uses '-RN' instead of '-TN'.
- For transcript features, different rules for isoform naming are used.
My questions are:
- Why are two separate suffixes used in the same product value for protein-coding isoforms?
- Why do the numbers used in the separate suffixes sometimes differ for the same isoform?
- Why use 'TN' for the protein-coding product suffix, 'RN' for the protein-coding transcript ID suffix, and 'XN' for transcript feature product suffixes?
Here are some examples from EGAPx runs (v0.4.1 and v0.5):
- product=heat shock protein 60A-like isoform 1-T1
- product=uncharacterized protein isoform 1-T2
- product=synaptotagmin 1 isoform 1-T3
- product=uncharacterized protein isoform 2-T3
- product=cramped chromatin regulator%2C transcript variant X5
Thanks for your help with this!
Activity
murphyte commentedon Mar 25, 2026
There are some background details to note:
For mRNA:CDS pairs, the final
productname is intended for human consumption and is based on the CDS. Differences in theisoform #counter mean the CDSes are different. The-T#suffix is added to indicate the mRNA transcript variant #. So a gene with 3 pairs of mRNA:CDS, where the first two have the same CDS, will be:isoform 1-T1
isoform 1-T2
isoform 2-T3
That structure retains the distinct transcript counters relevant for the mRNAs while allowing you to tell which CDSes are identical (if you know the naming convention).
For the
transcript_idandprotein_id, which later get transformed intoorig_transcript_idand the GenBank protein accession, they're constructed aslocus_tag-[R|P]#where R=RNA, P=Protein, and the # counter is the transcript counter. They need to be distinct for every feature, even for two identical CDSes, and this was a more typical convention.For non-coding transcripts from gnomon, there are two types:
neither feature type currently have their product names transformed from the original RefSeq style since they don't have paired CDSes to derive the name from. This could be a bit cleaner to at least switch the "X" to "T" for better consistency with the mRNAs. I think we can also add
orig_transcript_id, which isn't required for GenBank but would provide better consistency across the features.MonicaPoelchau-USDA commentedon Apr 3, 2026
Thanks for the explanations @murphyte! This helps.
I've talked through this with some of our users, and they agree that switching the "X" to "T" for non-coding transcripts would be helpful for consistency. That said, it's not a huge deal if it's too challenging to make the switch at this point.
I personally think adding
orig_transcript_idto misc_RNA and lncRNA features would be quite helpful for consistency, if they end up being indexed by GenBank.murphyte commentedon Apr 3, 2026
yep, I made a backlog ticket, but I don't know when we'll get to it. Thanks for asking -- I do agree clearing up the inconsistency is worthwhile in the long run.
MonicaPoelchau-USDA commentedon Jul 9, 2026
Hello again! I have another question about EGAPx isoform nomenclature.
We've noticed some cases where some isoforms receive an informative product name (e.g.
plasma membrane ascorbate-dependent reductase CYBRD1-like isoform 2-T2), but other isoforms of the same gene model don't (e.g.uncharacterized protein isoform 1-T1). The isoforms do not differ substantially from each other (89% ID at the transcript level in the example below). Do you know why that might be?Example:
murphyte commentedon Jul 9, 2026
@MonicaPoelchau-USDA - in the absence of orthology, individual isoforms of a gene get named based on their SwissProt best hit, as long as it's at least 50% query coverage. I suspect the hits are close to that coverage threshold such that one is above the cutoff and not the other.