Passing in filtered table.qza file for alpha and beta diversity analysis


I've filtered my table.qza file by metadata categories, but I'm getting an error when I try to pass it in to do diversity analyses. I'm not sure if I need to generate a new rep-seqs.qza file, but the sequences have already been denoised so that seems unnecessary. Any help would be greatly appreciated and thanks in advance!

Hello Irene,
We solved this problem by discovering that the table had been completely filtered out due to a misunderstood filtering in the SQL command. The error, however, is not great for figuring this out. @jairideout did we ever create an issue for this?

Hi @Irene_Zhang! I think you’re rarefying your table to an even sampling depth that is too high, which will cause all of your samples to be discarded. There is an associated issue to make the error message clearer (I think that’s what @Arron_Shiffer was referring to).

Can you check out what sampling depths are in your feature table (e.g. using qiime feature-table summarize), and compare those to the even sampling depth you’re suppling to core-metrics?

Hello Jai,
That is specifically it!

Hi Jai,

Yes, I was using the wrong keyword in filtering! I was trying to filter the table by metadata to retain samples where the metadata category “Origin” equaled “Sinus” and “Lung”. I used “Origin=‘Sinus’ AND ‘Lung’” instead of “Origin=‘Sinus’ OR ‘Lung’” and it filtered out all of my samples and resulted in an empty table! I was wondering if maybe adding some documentation on the use of AND vs OR to pick multiple values from a metadata category would make this clearer for us folks who have never used SQL?

Thanks for all your help @jairideout and @Arron_Shiffer!

Great, glad that solved the issue! We have an open issue to add more examples of SQL-based filtering to the filtering tutorial – I linked this forum post to the issue. Thanks @Irene_Zhang!