# unassigned taxa - oddly reversed between deblur and dada2 pipeline \*so confused\*

**URL:** https://forum.qiime2.org/t/unassigned-taxa-oddly-reversed-between-deblur-and-dada2-pipeline-so-confused/11696
**Category:** User Support
**Created:** [September 27, 2019, 11:48am UTC](https://forum.qiime2.org/t/unassigned-taxa-oddly-reversed-between-deblur-and-dada2-pipeline-so-confused/11696 "2019-09-27T11:48:51Z")
**Posts on this page:** 1
**Showing post:** 11

<div class="post-metadata">

### Author: ![Nicholas\_Bokulich](https://forum.qiime2.org/user_avatar/forum.qiime2.org/nicholas_bokulich/32/19937_2.png) [@Nicholas\_Bokulich](https://forum.qiime2.org/u/Nicholas_Bokulich)
#### Post date: [September 27, 2019, 3:41pm UTC](https://forum.qiime2.org/t/unassigned-taxa-oddly-reversed-between-deblur-and-dada2-pipeline-so-confused/11696/11 "2019-09-27T15:41:23Z")

</div>

> [@Cistron](#):
>
> here is the vsearch classifier output. This looks much more consistent, but why?

> [@Nicholas\_Bokulich](#):
>
> any chance were the samples at 8:00, 17:00, and 18:00 sequenced on a different run or different primers?

classify-sklearn cannot currently handle reads in mixed orientations — it guesses the orientation of reads based on the first 100 or so, and then processes them accordingly, but if reads in the opposite orientation crop up it becomes discombobulated. Those samples for some reason must have their reads in opposing orientations.

For your workshop I would recommend just excluding those samples, or use classify-consensus-vsearch to avoid this issue.

---

_[View the full topic](https://forum.qiime2.org/t/unassigned-taxa-oddly-reversed-between-deblur-and-dada2-pipeline-so-confused/11696)._