Available: Pre-trained classifier of V3-V4 (341F, 805R) region with gg_99

Community Contributions
1

Hi folks,

In case anyone is interested I have a pretrained classifier on the V3-V4 region using the following primers:
341F: CCTACGGGNGGCWGCAG
805R: GACTACHVGGGTATCTAATCC
The primer set is taken from Illumina's 16S library preparation protocol.

It was trained using the Greengenes 99_otu_taxonomy.txt using naive bayes method on qiime2017.12, its about 44MB.
If this is something you'd like to use let me know!
-Bod

17 Likes
2

Hi Bod,

I'm interested in the V3-V4 classifier you used. I have exactly the same primers pair in my data. I would also like to take a look on the commands you used, so I can try to reproduce it in my computer from the gg-13-8 full-length sequences database.

Thank you very much in advance!

ju4n

1 Like
3

Hi @ju4n_dc,

Here is a dropbox link to the classifier to download, let me know if you run into any problems. You can follow the exact commands within the provenance in the file to see what was exactly done.
Hope this helps!

-Bod

1 Like
4

Thanks Bod! It worked OK.

ju4n

1 Like
5

An off-topic reply has been split into a new topic: How to read provenance in an artifact file?

Please keep replies on-topic in the future.

6

2 off-topic replies have been split into a new topic: The scikit-learn version (0.19.1) used to generate this artifact does not match the current version of scikit-learn installed (0.20.2)

Please keep replies on-topic in the future.

7

A post was split to a new topic: V3V4 trim length and length parameters for extract-reads

8

2 posts were split to a new topic: How to obtain an updated V3-V4 99 greengenes feature classifier

9

5 posts were split to a new topic: Read length selection during extraction from reference databases

10

2 posts were split to a new topic: Should I use aligned or non-aliged representative sequences to train my own classifier

11

Hi Estaki, Can I have the trained classifier based on the 341F and 806R primer set? The Dropbox link is not working

Sayada

16

Hi @sakther1, Sorry for the slow reply. We're discussing how we can provide this and hope to have an update shortly. Note though that the full-length classifier should work almost as well, so that's a totally fine one to use too.

17

Hi @sakther1, My recommendation right now is to use the full-length classifier that is linked from https://resources.qiime2.org. There is not a very big performance difference between the region-specific classifiers and the full-length classifiers. An alternative though, if you do want to use a region-specific classifier, would be to train one using q2-feature-classifier. The files that @Mehrbod_Estaki used for training this one are available for download here.

1 Like