GAFs cannot be fully re-generated with the fields as they currently stand

[kltm]

The issue is that we don't currently index the input of GAF files as-is, rather they go through several layers of abstraction. By the time the information from the input file hits the index, the input (using an stc1 annotation as an example):

localization_dependent_on(GO:0031048)

has variously become:

 ["{\"relationship\": {\"relation\": [{\"id\": \"GOREL:0000009\", \"label\": \"localization_dependent_on\"}], \"id\": \"GO:0031048\", \"label\": \"chromatin silencing by small RNA\"}}"],

and:

chromatin silencing by small RNA

and:

GO:0031048

as well as the various closures.

After a little discussion, the solution will be to add one or more fields to the index for annotations that are:

• the entire input line
• the "hard" fields as they stand

[kltm]

Going to have to fix OWLTools for this first.

[kltm]

To clarify this for people searching, we're looking at column 16 (c16) in the GAF.