questions on docs for make-db --reference

[nick-youngblut]

According to the README docs:

The TSV file must contain at least the following columns: Accession, Genus, Species, Strain, ChromosomeLen, SeqPath, GFF. An easy way to do this is using the ncbi-genome-download tool and using our script (link here).

I have a few questions about this input. It would be great to have some clarification on the following:

• What is the format for SeqPath and GFF? Is GFF a path to a GFF file?
  • If so, it doesn't seem to follow the convention of SeqPath ("Path" in the name), so it's unclear.
• Is SeqPath a path to the assembly fasta file (nucleotide)?
• Can the fasta file be compressed (gzip or otherwise)?
• Can the GFF be compressed (gzip or otherwise)?
• What if the genome has no accession (e.g., a newly generate MAG)?
  • Can one generate a placeholder accession?
  • Is there a require format for the access string?

Also, it appears that the link is missing for link here.

[yk23]

Hi, sorry about this -- the software is still under development as we change some specifications for the paper we are submitting! (And Ive personally had a major life event recently and will resume work on Aug 1)

In the new version (previewable on the dataset-nt-jax branch), we provide a newer/more streamlined way to create the database, including the usage of MAGs. Changes to be merged into main sometime early August.

[yk23]

For an example (in the meantime), check out the dataset-nt-jax branch, in particular the klebsiella script 'examples/database/complete_recipes/klebsiella.sh' which will generate an index file as an example.

[yk23]

Finally, to answer some questions one by one: SeqPath and GffPath are both absolute or relative file paths to a fasta assembly file (e.g. nucleotide contig records im standard FASTA format parsable by BioPython)/GFF3 annotation file respectively. The software should accept GZipped versions for both (if it doesn't then thats a bug that's since been fixed on the experimental branch).

A placeholder accession ID suffices as long as it is unique (just to be safe, use a completely different format from ncbi). As long as your SeqPath is valid, the database indexer will automatically create symbolic links (in the unix sense) that the software will recognize.

A new as of yet undocumented change in the upcoming version is that the GFF3 is optional; the database will initialize (albeit with less meaningful metadata) even if it is not provided.