I’m interested in getting the difference (or changes) in taxonomy between 2 time points and then running ANCOM on those differences. This is shotgun data, so interested in species level changes. I have found a few topics on finding differences between time points in the forum, but not specific to taxonomy assignments. So essentially a pre-post type of design.
Is this the correct code to start? (found in forum)
qiime longitudinal pairwise-differences
–i-table table_relative-frequency.qza
–m-metadata-file mapping_file.tsv
–p-metric featuretable
–p-group-column Mucositis_Grade_Range1
–p-state-column Day_Group
–p-state-1 _min1
–p-state-2 7_13
–p-individual-id-column Personal_ID
–p-replicate-handling random
–o-visualization pairwise-differences.qzv
And if so, then how would one apply the ANCOM statistic after?
You have the correct command for looking at paired-sample changes across a time interval, and technically it could be performed on specific features in a feature table... but there are two issues to be aware of:
it sounds like you are interested in discovering which of many features change in relative frequency, not testing specific features. This action will only test specific features one at a time...
this action does not perform any transformation on the data, so on its own a paired Wilcoxon or t-test is not really an appropriate test for compositional data (e.g., relative frequencies). Technically you could transform the data with an appropriate method and then input those data to this action, but this is not possible in QIIME 2 at the moment on its own.
For paired testing with ANCOM it is easiest to test outside of QIIME 2 at the moment... this is a recent relevant discussion, including a simple way to do this with python/scikit-bio: