songbird question

Hello!
I have a question regarding songbird - more specifically how to change the songbird parameters to get the model to fit reasonably to my data.
(I want to explore whether there are certain features of the gut microbiome that are more associated with patients who gain wt vs lose weight).

This is the script I use:
qiime songbird multinomial
--i-table table.qza
--m-metadata-file metadata.tsv
--p-formula "Wt_category"
--p-epochs 20000
--p-differential-prior 0.5
--p-summary-interval 1
--o-differentials differentials.qza
--o-regression-stats regression-stats.qza
--o-regression-biplot regression-biplot.qza

The output images are attached. From what I understand the curves for each of the graphs should show decaying and plateau as close to zero as possible which is the not the case in my graph. I tried playing around the parameters by playing with the numbers for the p-differential prior and the p-epochs parameters but doesn't seem to help my case.
Can someone advise me what I can do to improve my model?

Thank you :sunny:

I'd generate some Q2 scores -- otherwise your plots look pretty good

2 Likes