The GFF3 format is well specified for describing sequence location and type in the first eight columns. The ninth column is left for specifying any remaining information. I would like to use GFF3 the encode the data typically produced by a genome annotator.
How should I go about this? Are there any conventions for encoding information such as protein product, EC number, and description?
•
written
by
Michael Barton •
There are no conventions outside of the GFF3 spec, though I would note that the reserved tag "Note" has historically been used for descriptions in GBrowse, so if you're going to be using GBrowse, it makes sense to put descriptions there. Generally I suggest that you encode the information in such a way that it will make it easier to use in whatever your downstream application for it is. Of course, you may not always know what that is, but that's my best advice.
Thanks for suggestion. Are there any common variants of GFF which include this type of information?
You mean that have protein information? Not that I can recall, though I can tell you that the GFF3 at NCBI has EC_number tags that violate the GFF3 spec (at least, they did the last time I checked). What is the end use that you have in mind?
Encoding additional genome annotation data in GFF3. Things like product and description.
Yes, but I meant, why do you want to do that? Who is it for, and what will they be using the GFF file for?
You'll see from the spec that there are a few conventions for column 9: key-value pairs, "reserved" keys, characters that require escapes and some conventions regarding case.
Apart from that, the key-value pairs are whatever you wish. This is in large part because GFF3 files are used by GBrowse. Track display in GBrowse often relies on small Perl subroutines in the config file, which read customised attributes from column 9.
You might want to look at the Bioperl script bp_genbank2gff3.pl, which goes some way towards "standardising" the conversion of GenBank format to GFF3.
Log in to answer this question.
