Filtering features at a percent threshold from different groups

[Alan_Chan (Alan)]

Hi all,

I am interested in filtering out features that make up less than 3% abundance within separate groups. How can I do this? There are two issues here. First, qiime feature-table filter-features sums across all samples and then filters but I want to sum them within certain metadata groups. Second, that same command doesn't take percent abundances only absolute counts, right? Would I have to convert the table to qzv then calculate the 3% threshold using the Total frequency shown in the qzv, then finally use that number for the --p-min-frequency parameter?

The only way I can think of doing this would be to separate each metadata group into their own feature tables, then convert each to qzv and visualize. Then manually calculate what ever the 3% threshold is for each feature table and use that to filter each separately. Finally, combine each feature table back into one. That seems like so much. Surely there must be a better way?!

[Nicholas_Bokulich]

@Alan_Chan

First, qiime feature-table filter-features sums across all samples and then filters but I want to sum them within certain metadata groups

Use filter-samples to split the table into groups, then use filter-features to filter on each group separately. Then merge back together with merge.

@Alan_Chan

Second, that same command doesn’t take percent abundances only absolute counts, right? Would I have to convert the table to qzv then calculate the 3% threshold using the Total frequency shown in the qzv, then finally use that number for the --p-min-frequency parameter?

Correct and yes that is an appropriate way to determine the total frequency for converting to a percentage.

@Alan_Chan

Surely there must be a better way?!

No, you have the right idea. What you are after is a fairly specific analysis, so unfortunately there is not an easier way to do this right now (that I can think of).