Older database support?

[wwood]

Hi,

I was keen to try out Metabuli, and was using an R207 pre-built database but now it seems that this has been superceded by the R214 one. For the sake of reproducibility, are old databases kept around?

One small suggestion also - it would be helpful if the GTDB db had the version coded directly into its filename, rather than simply gtdb.tar.gz. I realise that this info is in the README and once the tgz is unpacked, but would still be helpful so that the download link changes.

TIA, ben

[jaebeom-kim]

Dear Ben,

Thank you for trying Metabuli :)

Let me discuss about keeping older databases in the cloud server and write an answer about it later.

Regarding your suggestion, l think making the version information more visible is the point.
Renaming the DB link can be a solution, but it requires changing the source code and making new release/bioconda whenever the DB is updated.
And users also should update their Metabuli to download DBs from the new links.
So, I'm thinking about a different way to improve the version visibility.

Sorry for not giving direct answers.
I will post progresses about solving your concerns.

Thank you,
Jaebeom

[jaebeom-kim]

By the way, I'd like to recommend using R214 one.
In the R207 DB, K-mers from lowercased letters in a human genome were not extracted correctly.
So, it harmed Metabuli's performance for decontaminating human reads.

[wwood]

Thakns for the quick reply.

Renaming the DB link can be a solution, but it requires changing the source code and making new release/bioconda whenever the DB is updated.
And users also should update their Metabuli to download DBs from the new links.
So, I'm thinking about a different way to improve the version visibility.

There are other solutions. For instance metaphlan keeps an "mpa_latest" file which points to the versioned database files, and SingleM uses a library we made https://github.com/centre-for-microbiome-research/zenodo_backpack which operates through zenodo, where there is a concept of a DOI for a series and a DOI for a specific database version.

It would be great if each tool developer didn't have to spend time solving these problems and there was a broadly adopted, standardised solution.

By the way, I'd like to recommend using R214 one.
In the R207 DB, K-mers from lowercased letters in a human genome were not extracted correctly.
So, it harmed Metabuli's performance for decontaminating human reads.

Thanks - is there any other reason not to use the R207 one? My samples do not contain human reads.

[jaebeom-kim]

Thank you for providing great examples for the task!
If your sample doesn't contain human reads, R207 is totally fine.
R207 just has a smaller number of genomes.

[wwood]

Thanks @jaebeom-kim .

Congratulations on the metabuli paper.