How to convert Feature table [relative frequency] to a Feature table [frequency]

Hello!
First of all, thank you for the new release, it’s great!
Is there a way to convert Feature table [relative frequency] to a Feature table [frequency]? I found how to make the opposite only.
Thank you in advance

Hi @timanix,
Great question — there currently is no such conversion. Why?

  1. It does not really make sense — when you have relative abundance information you cannot convert that back to integer counts unless if you know the total counts. We could figure that out, of course, by looking at the original feature table, but then we need to ask:
  2. What's the point? Why convert to relative frequency and then back? I cannot think of any kind of steps, e.g., filtering, that can be performed on the relative frequency table, so this conversion adds no value. Just hold on to the original frequency table for additional analyses.

If you can describe a good use case, though, we could certainly consider adding to :qiime2:!

I hope that helps!

1 Like

Thank you for your answer! I don’t have much of the experience working with Qiime so I am a little bit lost and doing as much of analisys with my data as I can.
I was working with Qiime2 2018.6 and performed Longitudinal, ANCOM, Gneiss, then filtered my feature table (frequency) to the features, marked like a significantly changed according to the ANCOM output and created taxaplot. Then new release 2018.8 was issued and I performed additional calculations with updated Longitudinal, and got in the output filtered table, which includes only important features (according to the Longitudinal). I wanted to make a taxa plot with that table too but got a mistake, so I was thinking if I can convert that filtered table (relative frequency) just to frequency table and create a taxa plot.
As I already said, I am a new user and there is some workaround I don’t know yet.

Great question — that’s a brand new action in this release.

That file is just the FeatureTable[Frequency] table that was input, filtered by the FeatureData[Importances] file that that pipeline outputs, and converted to relative frequency. So you can just use the importances file to filter the feature table that was used as input, using qiime feature-table filter-features (just input the importances artifact as the “metadata” file to that command and it will do the rest).

I hope that helps!

2 Likes

Didn’t think about it! I forgot that I can use qza files as Metadata, thanks!

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