dbs used as targets : should we use all isoforms or only primaries?

[adf-ncgr]

@nathanweeks in your example blast result I noticed that a lot of the hits for Tifrunner seemed to be different isoforms of the same gene. I am not %100 sure but I think current practice has been to use the "primaryTranscript" forms- maybe @cann0010 can confirm or deny that? If this is true, just be aware of our somewhat confusing standard that when a dataset does not have a distinction between "primaryTranscripts" and the superset from which that is derived, then we just have the superset files (which will be equivalent to "primaryTranscripts" but not named as such). We should probably think of a way to make this less confusing.

[nathanweeks]

That appears to be the case for Tripal BLAST @ legumeinfo.org. SequenceServer now has cds_primaryTranscript BLAST targets (where they exist).

[StevenCannon-USDA]

Thanks for catching that. Yes: _primaryTranscript should always be the target. If there are annotation sets that lack this form, then we should make a change in the Data Store. I suppose we could use _primaryTranscript to indicate annotation files that lack alternate splice form - whether or not there is another annotation set WITH multiple splice forms. Your opinions?

[adf-ncgr]

pulling @sammyjava and @ctcncgr in regarding the handling of annotation sets with only one isoform annotation per gene. I don't recall if we had any reason for the approach we have taken so far, and I can't seem to find any github issue in which this was discussed. I am not sure whether any existing code actually relies on the current convention (ie where we do not have files called "primaryTranscript" in the case where there is no distinction between the full set of isoforms and primaries)

[sammyjava]

My practice has been to load _primaryTranscript if it exists; else load the one that exists. I load everything, of course. And I can easily change this practice, I do feel like we had a conversation on a Monday or something in which _primaryTranscript was recommended by @adf-ncgr but I may be making that up.

[adf-ncgr]

we did have a conversation, but I said I thought in the context of the mines we probably do want the full set of transcripts loaded,. My reason for this being that you will have the full set of isoforms loaded from gene_models_main, so why not also have their sequences? This would mean if we change the standard you'll have to look for _primaryTranscript files only if the others do not exist.

[sammyjava]

Sounds good, I probably misunderstood that original conversation. Will do in future builds.

[sammyjava]

Come to think of it, I used to think that gene_models_main.gff3 (the emphasis on "main") only included "primary" transcripts. So that was a fundamental misunderstanding, which I hadn't bothered to check by looking carefully at the GFFs. Anyway, all good.