Any update for the default database

[ZhangDengwei]

Hi there,

Thank you for the awesome tool. I notice that the default database was built two years ago; it would be great to update it consistently. Thanks!

[kdc10]

Hi,

We are not currently working on this, but would of course welcome anyone in our community to do so :)

[AlexandreThibodeauUdM]

Hello. How does one help? The database seem to be still be doing a good job, at the very least a better one then many other tools, but guess it could be updated every 5 years... lol.

How can I help? I have very basic skils but might be able to free some time to try to do something.

[jflot]

Same here. If there is somewhere a clear set of instructions on how to build a proper, new 16S database for Emu, I would be happy to try doing so and share the result. However, I don't want to risked making a botched attempt then papers from many authors would have to be retracted once it would become apparent that my reference library was not made properly...

[kdcurry]

The current emu database is a summation of the rrnDB and NCBI 16S. This could be repeated with updated versions, following the steps in the manuscript: https://www.nature.com/articles/s41592-022-01520-4. The most tedious process will be making sure the taxids from both sets of sequences match the provided taxonomy (we used NCBI, but be wary that this changes daily). Once this is done, the build-database function could be used to ensure the format aligns with emu's abundance script. Other curated 16S databases are also welcome, although we have found that the large size and incomplete taxonomy holes in bigger 16S databases such as greengenes, silva, and RDP did not produce more accurate results in our previous tests. These databases tended to yield more representatives for the same species, yet Emu would benefit from a database with a larger range of species representation and fewer sequences per species.

The new database should then be validated against existing test datasets. We used a dataset from CAMI II (https://www.nature.com/articles/s41592-022-01431-4), although I believe any defined benchmarking metagenomic dataset would be fine.