Hello,
I am trying to understand my ANCOM-BC output. In the metadata column I have total 9 different combinations of treatments. When I run the analysis, it gives me links for 9 different charts with blue and orange lines. Please see the attached image for one treatment only. According to my understanding , this plot is against all other treatments?
Also, I want to show only family-level taxa on the plot, how can I exclude the other taxa-levels.
Thank You in advance.
SZ
Hi @Szahra,
2 quick things will adress the readablity of the plot:
re-run da-barplot but
- Use the parameter --p-level-delimeter ";" . This will split taxonomic strings and just give you the last taxon grouping.
- set a significance threshold. This will drop any features that doesnt have a p-value of greater than your threshold which will reduce the size.
Not quite. Your 9 plots are 1 treatment compared to your reference. If a reference wasn't provided then ancombc chose the first group alphebetically in your metadata column.
You are probably looking to run qiime2 taxa collapse and collapsing at the family level.
https://docs.qiime2.org/2024.5/plugins/available/taxa/collapse/
Can you please provide your command for ancombc and da-barplot so i can get a better idea of what is happening?
Hope this helps!
4 Likes
Thank You for the reply. Yes I have used taxa collapsed table. My objective is to find differential abundant taxa in each treatments against all other. I have 3 genotypes (A, B, C) with 3 Amendments for each genotype (lets say : Control , 1 and 2 ). So, In metadata column for "treatment" I have total 9 combinations).
Here is my code :
1- qiime taxa collapse --i-table table.qza --i-taxonomy taxonomy.qza --p-level 6 --o-collapsed-table collapsed_level6.qza
2- qiime composition ancombc --i-table collapsed_level6.qza --m-metadata-file metadata.txt --p-formula Treatment --o-differentials differentials.qza
3- qiime composition da-barplot --i-data differentials.qza --o-visualization differential_barplot.qzv
Do I need to split the data for each reference (Control ) for Amendment 1 and 2 ?
Thank You
Hi @Szahra,
Jumping in for @cherman2 here! I'm going to provide you with some additional information that should indirectly answer your question below.
Each time you run ANCOM-BC, you're selecting the reference level(s) for differential abundance calculations - this is essentially setting which group(s) the differential abundance should be with respect to. So for your Treatment groups, you'd pick the group you'd like to set as the 'intercept' or reference - and in the results, the LFC for each group shown will be with respect to that. If no group is selected, the default intercept (i.e. --p-reference-levels) will be the first group in Treatment in alphabetical order. As an example, if you had the following groups within Treatment:
FMT
antibiotics
none
The default reference level chosen would be antibiotics. You could specify a different group by passing your preferred group into --p-reference-levels. If you'd like to see the LFC with respect to each group within Treatment, you'd need to re-run ANCOM-BC for each group, changing the reference level accordingly each time.
No additional data modification/preparation should be required for you to do this - but I would reiterate @cherman2's suggestion of utilizing the --p-level-delimiter ";" argument in order to improve the readability of your barplot.
Hope this helps! Cheers 
3 Likes
Okay , Thank You
@lizgehret for the detailed explanation , I understood it now. Also, after using `--p-level-delimiter ";" and significance arguments, my graphs are clear and concise. Thanks to @cherman2.
SZ
2 Likes
This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.