# Pretrained SILVA 138.1 classifiers (weighted and uniform), reference dbs, and habitat weights

**URL:** <https://forum.qiime2.org/t/pretrained-silva-138-1-classifiers-weighted-and-uniform-reference-dbs-and-habitat-weights/22627>\
**Category:** Data resources\
**Tags:** feature-classifier\
**Created:** [March 31, 2022, 1:56am UTC](https://forum.qiime2.org/t/pretrained-silva-138-1-classifiers-weighted-and-uniform-reference-dbs-and-habitat-weights/22627 "2022-03-31T01:56:21Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![BenKaehler](https://forum.qiime2.org/user_avatar/forum.qiime2.org/benkaehler/32/1693_2.png) [@BenKaehler](https://forum.qiime2.org/u/BenKaehler)\
**Post date:** [March 31, 2022, 1:56am UTC](https://forum.qiime2.org/t/pretrained-silva-138-1-classifiers-weighted-and-uniform-reference-dbs-and-habitat-weights/22627/1 "2022-03-31T01:56:21Z")

</div>

Pretrained SILVA 138.1 classifiers for use with `q2-feature-classifier` are now available on [Zenodo](https://zenodo.org/record/6395539).

You can download classifiers for full-length 16S or 515f-806r (V4) trimmed reads. There are uniform classifiers and classifiers weighted for animal-distal-gut, animal-surface, animal-secretion, soil-non-saline, or animal-corpus EMPO 3 habitat types. There are also weighted classifiers for human-oral or human-stool samples.

There are also "average" weighted classifiers that use weights averaged across 14 EMPO 3 habitat types. We [found](https://doi.org/10.1038/s41467-019-12669-6) that those generally outperformed uniform (ie. unweighted) classifiers. The habitat-specific classifiers were even better.

If you have samples from a range of other EMPO 3 habitat types, you can train your own custom weighted SILVA 138.1 classifier using weights and reference databases available at [readytowear](https://github.com/BenKaehler/readytowear).