Differential abundance using AMCOM

Hi Everyone,

I am trying to explore any differentially abundant taxa in my study using ANCOM. Especially, I am interested in investing the difference in MOA1 vs MOA2 vs MOA3 (column names in metadata file).

16S_OR_WA.qzv (1.2 MB)

I used the code and have the following question-
[qiime feature-table filter-features --i-table filtered-table-4.qza --p-min-frequency 50 --o-filtered-table table_7.5k_abund.qza

qiime composition add-pseudocount --i-table table_7.5k_abund.qza --o-composition-table table7.5k_abund_comp.qza

qiime composition ancom --i-table table7.5k_abund_comp.qza --m-metadata-file 16S_OR_WA.tsv --m-metadata-column MOA1 --o-visualization ancom_MOA1_qzv ]

filtered-table-4.qza (4.8 MB)

ancom_MOA1.qzv (1.9 MB)

The analysis took unexpectedly longer more than a day to run- is this usual? and surprisingly, I did not find a single significant taxon. These results made me think maybe I am not doing it right.
Can someone point me in a right direction or have any thoughts about it?
Thank you in advance for your time.

I further used the "Moving Pictures” tutorial - Multiple Interface Edition" to run the analysis.

qiime feature-table filter-samples --i-table filtered-table-3.qza --m-metadata-file 16S_OR_WA.tsv --p-where '[period_label]='"'"'post12'"'"'' --o-filtered-table post12_table.qza

qiime composition add-pseudocount --i-table post12_table.qza --o-composition-table comp_post12_table.qza

qiime composition ancom --i-table comp_post12_table.qza --m-metadata-file pooled_16S_0328.tsv --m-metadata-column MOA --o-visualization ancom_post12_qzv
and I have the following error.

filtered-table-3.qza (4.8 MB)

16S_OR_WA.qzv (1.2 MB)

Any feedback will be highly appreciated.

Thank you

Hi @shaista_karim,

The two issues are unrelated. In the second, the issue is with your metadata and your feature table not lining up. Double check the table you're passing and your metadata.

The first tells me you're testing a lot of features, and that's why it takes so long to run. My guess based on your provenance is that you have a lot of singletons and not a lot of samples. My recommendation would be to be relatively stringent. Personally, I tend to try for 10% of samples to have a feature at whatever I've defined as "present". I would also check your beta diversity: do you see a significant difference in any of your metrics?


hi @jwdebelius thank you for giving feedback.

I double-checked everything- it is the metadata I need to pass. I used qiime feature-table summarize --i-table table.qza --m-sample-metadata-file WA_16S_0331.tsv --o-visualization table.qzv
and it turned out just fine.
table.qzv (2.4 MB)
This explain it is not the metadata issue as far as I understand.

I am using Parkinson’s Mouse Tutorial — QIIME 2 2022.2.0 documentation for differential abundance analysis.

The steps I am using is
qiime feature-table filter-features --i-table table_2k.qza --p-min-frequency 50 --p-min-samples 4 --o-filtered-table table_2k_abund.qza
qiime composition add-pseudocount --i-table table_2k_abund.qza --o-composition-table table2k_abund_comp.qza

qiime composition ancom --i-table table2k_abund_comp.qza --m-metadata-file WA_16S_0331.tsv --m-metadata-column MOA --o-visualization ancom_MOA.qza

table.qzv (2.4 MB)

WA_16S_0331.qzv (1.2 MB)

I think setting up min frequency and min sample might have cause the problem. I tried couple of many approaches but not good luck yet.

Thank you in advance.

@jwdebelius looks like the command does not like the blanks in metadata. I then use Filtering data — QIIME 2 2022.2.0 documentation to filter out my MOA column in metadata which as MOA1, MOA2, and MOA3 data using qiime feature-table filter-samples --i-table table.qza --m-metadata-file WA_16S_0331.tsv --p-where "[MOA] IN ('MOA1', 'MOA2', 'MOA3')" --o-filtered-table MOA_filtered_table.qza and rerun everything again. did not work either :frowning:

MOA_filtered_table.qza (995.0 KB)

ancom_MOA.qza.qzv (792.0 KB)

Hi @shaista_karim,

I'm glad you sorted out the missing metadata issue; this varies between commands and can be somewhat hard to predict.

I'd like to repeat my earlier questions, though:

  1. Do you see a difference in beta diversity?
  2. Do you think you're filtering enough?

The parameters from the PD mouse tutorial may not be appropriate for your data. That's a 16 sample dataset (so 4/16 = 25% prevalence) and the average average depth is around 2500 sequences/sample, so those parameters might not be appropriate for you. I would suggest looking at your filtering options in the q2-feature-table documentation directly since there are some functions there which are not covered in the tutorials.


HI @jwdebelius thank you for getting back.

Yes, I saw differences in my beta diversity- because of that I thought I should explore what taxa are deferentially abundant.

I will try more stringent approach and explore more filtering options you shared and get back

1 Like

Hi @jwdebelius -- Thank you so much, my problem got solved.

ancome_period_label.qzv (526.2 KB)


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