When using qiime diversity core-metrics-phylogenetic, significantly different results are generated

Hi, about a month ago, I used the "qiime diversity core-metrics-phylogenetic" plugin on a feature table and successfully obtained results. Later, I noticed that one sample (ID: NT-55) was not displayed correctly in the PCoA plots, although this sample still exists in the feature table.

To understand why this sample point disappeared, I re-ran the command to generate the distance matrix. To my surprise, the newly generated plot was significantly different from the original one (I didn't make any modifications to the command line). I repeated this command three times, and the new results are highly consistent with each other.

I'm not sure what caused such significant differences or why you couldn't replicate the results from the first run (I used the --quiet option, so I don't know what the output was when it ran). The three attachments are the files used for the analysis (feature table, metadata, and phylogeny tree), the results of the first diversity analysis, and the most recent analysis results.
featuretable_metadata_tree.zip (1.2 MB)
initial_core_metrics_results.zip (5.5 MB)
new-core-metrics-results.zip (3.2 MB)

Thank you in advance to anyone willing to help me.

Good evening,

Thank you for posting those artifacts. I opened them up and compared their provenance.

The input data is exactly the same. The DADA2 settings are different!

This would absolutely cause the tables, distance matrixes, and emperor PCoA plots to change.

Any ideas about how this could have happened?


Hi Colin,

Thanks for the help. Yes, I did try different parameters for DADA2 to get the best-denoised results. Could you please tell me how to tell the difference in DADA2 settings based on the provided artifacts?


1 Like


Go to https://view.qiime2.org
Drag your artifact into that window
Provenance > click on graph node > Action Details

(this is from the initial core metrics file you posted)

1 Like

That solved my question. Thank you!

1 Like