SILVA trained version for V3-V4 amplicon using 341F/785R primers

General Discussion
1

Hello everyone,

I’ve recently started using QIIME2 for a 16S soil microbiome analysis. After running the taxonomy classification, I noticed that I obtained a very low number of ASVs (around 1,000).

Someone pointed out that I might have used an incorrect classifier. Indeed, I used the SILVA-138 classifier trained for the V4 region (515F/806R), whereas my sequences were amplified using the V3–V4 region (341F/785R).

I attempted to train a classifier myself, but the process was terminated by the server (something I’ve read others have experienced as well).

I’m wondering if anyone already has a SILVA-138 classifier trained for the V3–V4 region (341F/785R) and would be willing to share it.

Thank you very much in advance!

2

You are likely running out of memory. I'd suggest requesting at least 32 GB - 64 GB of RAM. Also, dereplicating the data prior to classification will also help save on RAM usage.

Using a pre-made full length classifier should be sufficient, though, it does use more RAM when using it to classify sequences.

If you search the forum, you might come across some links to premade V3V4 classifiers, but they may not be compatible with your version of QIIME 2.

1 Like
3

Hello @Minely, welcome to qiime2forum.

I just trained a V3-V4 (341F-806R) SILVA-138.2 classifier on qiime2 2025.4, following the RESCRIPt tutorial [Processing, filtering, and evaluating the SILVA database (and other reference sequence data) with RESCRIPt].
I published it on GitHub along with the fit and taxonomy evaluations visualizations.
I'm no expert at all in classifier-training, but you can check it out if you think it could help you.
The classifier repo is GitHub - tripitakit/qiime2classifiers.
Hth.

3 Likes
4

Thank you so much!! I really appreciate it.