ANCOM results - significance of results

User Support
1

Hello :slight_smile:

I performed ANCOM on my data looking at two clinical groups (weight gain vs weight loss), and I'm not sure if the taxa (at genera level) are statistically different between the groups, as the W value for some is 0. Please see attached qzv file.

Note: the feature table is filtered so that any feature that is not present in ~10% of the samples is not included.

Can you help me interpret the results? And if there is anything I can do differently?

I've read this forum post already but not sure I quite understand it and how I can apply it to my data.

Using qiime2-2022.8 version

Command I'm using:

qiime composition add-pseudocount
--i-table filtered_table.qza
--o-composition-table filtered_table.qza

qiime composition ancom
--i-table filtered_table.qza
--m-metadata-file metadata.tsv
--m-metadata-column weight
--o-visualization wt_ch

wt_chg.qzv (536.5 KB)
ge.qzv

2

Hi @rnasrah,

  1. How do you currently interpret the results?

  2. In that post , they are talking about setting a threshold for what W score you need to get in order to consider them significant. @jwdebelius says she tries to have her W score be 70% of her comparison in order to consider it a significant finding.

  3. You could collapse your feature table to Family or Order, you might see more significant patterns.

Hope that helps!
:turtle:

3

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.