Filtering OTU table by abundance <0.1%

Hi all,

I have an OTU table containing my samples and several OTUs.

I would like to conduct an abundance-based filtering where I end up with an OTU table that includes only those OTUs that are present at above a confident level of detection (0.1% relative abundance).

So basically filtering out OTUs <0.1% of total reads.

Could anyone please help me? I’m aware that there is a way of doing this in QIIME1 but I am struggling to find something similar in QIIME2.


Hi, @thextramile
Depending on your analysis, abundance-based filtering may not be necessary in QIIME 2. This post might provide useful context. If you have a specific downstream need for abundance-based filtering, please provide details so someone with more experience can weigh in on how to approach your specific need.

Side note, questions about how to perform analyses should be filed in General Discussion, not User Support. Please take a look at November’s Changes to the Forum if you have any questions.

Thanks, and all the best,

1 Like

Thanks for your prompt reply @ChrisKeefe. I am new to QIIME and therefore don’t have much experience. However, from what I’ve read so far several people tend to filter their OTU tables to get to a smaller number of OTUs at the end of the analysis (around 500).

I currently have 6000 OTUs and if I open the final OTU table I can clearly see that several OTUs are only present in a few samples and have very low number of reads (sometimes even 0).

Filtering at 0.1% is the approach that I’ve been recommended to take by other researchers/colleagues in my field.

Is this something that cannot be done in QIIME2?

Thanks for your help! :slight_smile:

1 Like

@thextramile, many QIIME 2 users denoise with DADA2 or Deblur, so your approach is a little less common than it used to be. You can use qiime feature-table filter-features to filter by abundance - you may have to do some math to convert your desired relative frequency to an appropriate frequency threshold. As discussed in the post I referenced above,

the filtering we currently support operates on FeatureTable[Frequency] , not FeatureTable[RelativeFrequency] .


This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.