Some documentation I found suggests that a GenBank GI number will change each time the sequence changes - even if only one base is affected. The Accession number, on the other hand, remains the same.
However, the accession number is usually qualified with a version number suffix, e.g. "GL000191.1". As far as I know, this version number also increments each time the sequence changes.
Does this mean that the relationship between GI number and the Accession.Version pair is one-to-one, and so either would be equally suitable as a unique identifier for a sequence?
Thanks.
Yes, that's right. Accession.Version is probably preferred by humans, and GI number by machines.
Sorry, I can't resist :-P http://lmgtfy.com/?q=NCBI+gi
@Pierre, hmmmm, while I can see the funny side, I had read that page, but I still wasn't 100% clear if the relationship was always one-to-one... :) e.g. I once thought that the version number would increment if the sequence changed OR the meta-information of the sequence changed, e.g. the gene symbol. I now know that isn't the case now, but I thought maybe GI would increment on a meta change....