Update fixup_gene_name() to select the lowest expect allele, rather than just *01 when no allele provided.

[bbimber]

@sschattgen: this is a suggestion related to our email in December. This probably would help the situation with Rhesus macaques, where some segments do not have an 01 allele, but they do have 02.

[phbradley]

Hi there-- thanks for this new code! Quick questions

• is the natsort package guaranteed to be installed?
• would this also work with sorted instead of natsorted? Seems like the *01 allele should generally be first as long as the allele suffix has length 2, and is numeric,... But maybe there's a case I'm not thinking of?

[bbimber]

@phbradley : thanks for the reply. to those questions:

1) You probably have a point on natsort. Python isnt a language I know that well, and I mistakenly thought python could detect imports. I added natsort to setup.py.

2) Yes, I bet in virtually all cases sorted() would work instead of natsort. I chose natsorted() because if a given gene has an 10 allele, then this would be chosen preferentially over a 02 allele. Granted, this is unlikely, but it seemed better than put a known issue into the code.

[phbradley]

OK, thanks for the reply. My preference would be to use the built-in function sorted, to reduce dependencies and prevent breakage with existing python environments. It seems to give a reasonable result in this case:

In [9]: sorted(['TRBV19*01','TRBV19*02','TRBV19*10'])
Out[9]: ['TRBV19*01', 'TRBV19*02', 'TRBV19*10']

Does that sound alright?

[bbimber]

Interesting, I did not expect python to behave that way. I agree sorted() makes sense over natsort(). I just made those changes

[phbradley]

Great, thank you!