Support for INSDC Sequence records (i.e., Genbank/EMBL format)?

[laserson]

Any thoughts about supporting the INSDC data model as well? This would be for the data in genbank/embl format. This corresponds to the SeqRecord data type in biopython. Apologies if this questions is totally redundant/annoying.

[fnothaft]

I think that'd map well to @heuermh's bigdatagenomics/bdg-formats#83, no? If so, SGTM!

[laserson]

Yes, but it would extend a ways past it. What @heuermh currently implemented is similar to Biopython's Seq object. The SeqRecord object includes one of those, but also includes per-letter annotations (e.g., quality scores), dbxrefs, and a list of Feature objects. Basically, the sequence and all the annotations on top of it. Is there any appetite to develop that?

[fnothaft]

Oh, that's kinda cool. I might say that we could add the per-base annotations and the dbxrefs, but I would leave out the Features, since we can reconstruct that from a join. @heuermh thoughts?

[heuermh]

I was thinking Sequence and Feature as currently defined, and then a SequenceAnnotation or similar for the list of features and other stuff that goes in a Biopython SeqRecord or Biojava RichSequence. Didn't the GA4GH recently propose something for this?

[fnothaft]

Nice! I like that proposal.

[heuermh]

@laserson What would you like to use as the specification/reference model for this?

I've written Biojava 1.x (biojava-legacy) to/from Sequence/Slice/Read conversions here, similar could be done for Biojava 5.x on HEAD (biojava), and it wouldn't take much to use those libraries to support Genbank/EMBL formats in addition to FASTA and FASTQ.

Since Biojava is LGPL version 2.1, there is a question of where this code should live. There is also a question of the proposal at https://github.com/bigdatagenomics/bdg-convert.

[laserson]

I don't have experience with Biojava but probably a good place to start?

[heuermh]

A good place to start, sure, but some of the code is 15 years old. Is there a more up-to-date specification or reference?

There's this for the feature table
http://www.insdc.org/files/feature_table.html

and an annotated example Genbank record
https://www.ncbi.nlm.nih.gov/genbank/samplerecord/

I think I'm with Chris Fields in that, well, good luck
https://www.biostars.org/p/165727/

Given a Genbank record, which bits would be important to you?

[laserson]

Yeah, I'm also aware of how unspecified the Genbank format is.

For me, the obvious things to add are Features on top of the sequence. Also nice would be the per-letter annotations. And also the ket-value pairs you can add (is this called "qualifiers" or something like that?)

[heuermh]

Fixed by https://github.com/heuermh/biojava-adam, which is in the process of being transferred to the Biojava organization.