Metafile.tsv (1.8 KB)
paired-16SV1-filtered-table.qza (79.2 KB)
ancom-genotype-diff-filtered.qza (28.2 KB)
ancom-barplot-genotype-filtered.qzv (210.2 KB)
Hi all, when running ANCOMBC on of my sample groups gets dropped with no error or explanation. My experiment is using several different genotypes each exposed to PHE or control, each with 3 replicates (so a total of 6 per genotype with 3 treatment and 3 control). In all other analyses (eg PERMANOVA and beta diversity PCoA plots) all genotypes are displayed/included. However in ANCOMBC, (12.5_1) is not included in the differentials artefact (checked by unzipping the qza). The only difference with this genotype is that one of the treatment replicates has a low number of reads. I have tried filtering the feature table to include fewer genotypes and still have the same issue and also adjusting parameters in the ANCOMBC command to be more permissive, but still have the same issue. I have attached my metadata, and filtered table, filtered differentials and filtered da-barplot (as I would rather not share my entire feature table, so I have just shared the filtered version and artefacts made from this). Here is the command I used
Hello @FGigl,
Welcome to the forums! :qiime2:
Thank you for your detailed explanation of your experimental design.
Yes, only genotype7.5_4 and genotype88_1 made it to the ANCOM step...
Looking at that graph, I believe ANCOM is working as intended, but the filter is not!
I checked the provenance with https://view.qiime2.org to find your filter settings:
where: "[genotype] IN ('88_1', '12.5_1', '7.5_4')"
These seem to match the values in your metadata, so that's good!
Out of curiosity, what happens when you ANCOM-BC again on the full data, just skipping the filter step?
Hello @colinbrislawn,
thank you for your answer. Also with skipping the filter step 12.5_1 is still missing.
I initiallz wondered if it could be a mismatch between the metadata and the sample ids, but as 12.5_1 show up in the other visualisations (e.g. beta diversity PCoA) that doesn't seem to be the case. I would rather not share my entire dataset's denoised feature table or ancom differentials result as the data is unpublished (hence uploading the filtered files in my last post) but would be happy to share directly in a private message if you have time to have a look? This has really got everyone in my office stumped!
Yes! Feel free to DM me or one of the Staff members (say, @lizgehret ) with your files and we will take a look.
If you can make a mock data set that reproduces this error, even better!
This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.