Partially or completely support the SELENON gene.

[ifokkema]

Adapted from NM_020451.2:

The SELENON gene encodes for a selenoprotein, containing the rare amino acid selenocysteine (Sec). Sec is encoded by the UGA codon. The 3' UTRs of selenoprotein mRNAs contain a conserved stem-loop structure, designated the Sec insertion sequence (SECIS) element, that is necessary for the recognition of UGA as a Sec codon, rather than as a stop signal. A second stop-codon redefinition element (SRE) adjacent to the UGA codon has been identified in this gene (PMID:15791204). SRE is a phylogenetically conserved stem-loop structure that stimulates readthrough at the UGA codon, and augments the Sec insertion efficiency by SECIS.

An example of where this currently fails is NC_000001.10:g.26140612_26140626delinsT. All endpoints strangely enough return an p.= prediction (not even a p.(=)). Obviously, it's not easy to recognize selenoproteins, but perhaps some steps can be made.

• [x] Returning p.(=) instead of p.= would be an improvement already.
• [ ] Since the transcripts are annotated as selenoproteins, it's also possible to annotate these transcripts internally. There could be an online resource perhaps that lists these genes or transcripts, otherwise the NM annotation could be parsed? With this annotation, you can add a warning to the output, or choose to support these transcripts alltogether.
• [ ] If this SRE and/or SECIS element would be detectable by you, you might be able to determine if this TGA will cause a readthrough or not. Otherwise, trusting the NM's given CDS length above the sequence would also work. If the CDS given is divisible by three, the CDS is longer than the stop that you find, and the CDS length coincides with a stop in the sequence, you can incorporate a readthrough aminoacid where earlier stops are found, and continue normally. If there are other issues, just throw a p.?.
• [ ] If you're having trouble finding where to store this annotation, a hack could be to store UGA instead of TGA where you have determined a readthrough to be, and make UGA map to the Sec aminoacid. That's a bit of a hack perhaps, but it could work if you're looking for ways to do it.

[Peter-J-Freeman]

To remind myself, I think that p.(=) shokiuld always be used instead of p.= . We have thought about this before, and I think it does make sense to implement this change. I will do this. now.

There is an alternate translation Alphabet used by Biopython that deals with this issue in theory. Shouldn't need a hack. I'm going to leave this for now until after the next stable release and start putting ideas together here for us to discuss.