ANCOM visualization

Hi everyone, I'm new to QIIME2 though I've been using QIIME for a long time. One of the main reasons I finally made the transition was that I wanted to use ancom to explore my otu/taxa tables in a way that was independent of relative abundances. When I run ancom get the following visualization artefact- a few statistically significant features, but no volcano plot. Will ancom not generate a plot when only a few otus are significantly enrichedl6-ancom-Type.qzv (39.0 KB)

Hi Everyone,

I’m having the same issue here when analyzing my data.

Thanks,
FS.

@duderostome looks like there is a problem in generating the volcano plot. But it is hard to tell what the problem is without looking at the command / files.

Do you have the command / files used to generate this plot readily available?

Sure, here’s the relevant ancom commands, I’ve run several filtered datasets through a bash loop (hence the variables).

qiime feature-table filter-samples --i-table otu-table.qza --m-metadata-file Combinemaps2.txt --p-where "Site='$i'" --o-filtered-table "$i"-table.qza

qiime composition add-pseudocount --i-table "$i"-table.qza --o-composition-table comp-"$i"-table.qza

qiime composition ancom --i-table comp-"$i"-table.qza --m-metadata-file Combinemaps2.txt --p-transform-function log --m-metadata-category Type --o-visualization ancom-Type.qzv

qiime taxa collapse --i-table "$i"-table.qza --i-taxonomy All2_otus_tax_assignments.qza --p-level 6 --o-collapsed-table "$i"-table-l6.qza

qiime composition add-pseudocount --i-table "$i"-table-l6.qza --o-composition-table comp-"$i"-table-l6.qza
qiime composition ancom --i-table comp-"$i"-table-l6.qza --p-transform-function log --m-metadata-file Combinemaps2.txt --m-metadata-category Type --o-visualization l6-ancom-Subject.qzv

Hello Matthew,

This might not make any difference, but on your final command, I noticed that the variable was not included in the output:
--o-visualization l6-ancom-Subject.qzv

I expected something like:
--o-visualization "$i"-l6-ancom-Subject.qzv
But maybe I’m missing something…

Colin

Thanks for your response, I did omit the variable name from the output, I’ll add it in later when I get past experimenting with scripts. Each of the output files is actually being written to a separate folder with the variable name, so this isn’t an issue. Sorry this wasn’t clear from the commands.

Hi @duderostome, I’m not sure if I’m able to completely replicate your issues. Could you do the following

  1. What is your qiime2 environment? There have been some changes with the metadata, in the last few versions. It is unlikely, but it may be possible that there is something weird going on with the reading of the metadata table.
  2. Do you get an error when you add the --verbose flag at the end? If you were to run
qiime composition ancom --i-table comp-"$i"-table-l6.qza --p-transform-function log --m-metadata-file Combinemaps2.txt --m-metadata-category Type --o-visualization l6-ancom-Subject.qzv --verbose

Do you see any errors?

1 Like

Good advice. Here’s the output (for one set of samples):

Looks like it could be a NumPy rounding error?

/Users/rogersm/Downloads/yes/envs/qiime2-2017.12/lib/python3.5/site-packages/scipy/stats/stats.py:2968: RuntimeWarning: invalid value encountered in double_scalars
f = msb / msw

Note that the new Metadata changes haven’t made it into a QIIME 2 release yet – they’ll be included in the upcoming release (2018.2).

1 Like

Sorry - forgot the environment: qiime2-2017.12

Just to check up on this. Are you and @mortonjt working through this via DM/email? I would feel bad if this topic got lost!

@mortonjt and I worked on this for a bit - thanks for your time @mortoenjt! But in the end it remained unresolved. Is anyone else reporting this issue at the moment?

@duderostome sorry for the late response.

It looks like the problem has to do with the total genus abundances.
The f_test fails when all of species abundances are the same. In this case, there are some genera that are basically all pseudocounts. So if you want to have the visualization, I’d recommend trying to filter out some of those low abundance genera.

1 Like

This worked - thanks for all your time looking into this Jamie!

2 Likes

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