search

airr-community / airr-formats

PLEASE SEE airr-standards FOR FURTHER DEVELOPMENT: https://github.com/airr-community/airr-standards
MIT License
1 stars 2 forks source link

How do we encode `cdr3_aa` in cases of out-of-frame sequences? #24

Closed laserson closed 7 years ago

laserson commented 7 years ago

Shugay:

We usually translate in both V->J and J->V directions and mark the incomplete codon with '_'

javh commented 7 years ago

I think the official denotation of an unknown amino acid is X. My feeling is that - is probably appropriate as well, if derived from a gap rather than an ambiguous nucleotide.

http://www.bioinformatics.org/sms2/iupac.html

schristley commented 7 years ago

By out-of-frame do you mean that the rearrangement is non-productive? Or are you referring to the uncommon case where indels puts the rearrangement out-of-frame but it still happens to be productive? In the case of a non-productive rearrangement, I don't think it is necessary to translate CDR3.

laserson commented 7 years ago

My feeling is that we don't mandate this for now, and it can be a use-at-your-own-risk situation? Though I'll let @mikessh chime in here.

nishmm commented 7 years ago

@schristley Doesn't the tool that determines if its productive or unproductive has to translate it and determine unproductive (due to stop codon) in a particular frame?

schristley commented 7 years ago

@nishmm There are other ways for a rearrangement to be marked unproductive, but yes one of steps would be to check CDR3 for a stop codon after translation. The discussion we had in iReceptor the other day sounds like it is specific to MIXCR (other tools?), it might conclude that a CDR3, which has a length not divisible by 3, is actually valid with the missing/additional nucleotides assumed to be sequencing errors. This was how I understood Bojan's description. Other tools may be different.

laserson commented 7 years ago

This seems to be tool-dependent to some extent, and probably addressed using custom fields. Not sure there is any actionable thing here, so I am closing.