ANCOM for three treatments

Hi qiime2 experts,

Based on the ANCOM analysis, three microbial taxa (Gemmatimonadota, Latescibacterota and Methylomirabilota) were the significant taxa selected for the three treatments (red, blue, and green boxplots). Can we safely say that at least two of the treatments are significantly different?

image

Looking forward to hear from everyone.

Thank you.

Cheers,
Yit Kheng

Hi @Yit_Goh,
Welcome to the QIIME2 Forum:qiime2:
Disclaimer: I am not a statistician so I would suggest doing your own research into my answers.

I believe that you can not assume significance.
First, You can’t claim significance from your Ancom results because does not test for significance in its results.

Second, I am assuming that you made this boxplot based on the table that Ancom produces(without more info this is what makes the most sense to me). That Ancom table is absolute counts, not relative frequency. Boxplots based on those absolute counts are not really representative of your data.

Last, this box plot is a great quantitative representation of your data (with the above caveat) but the only way to be able to claim significance is through a qualitative analysis of your data. Statistical tests that give p-values will allow you to make claims of significance but not just looking at a boxplot.

Hope that helps!
Chloe :turtle:

3 Likes

Hi @cherman2,

Thank you for the answers related to the use of significance for ANCOM results. You are right, the boxplot was generated based on the table produced by ANCOM (composition table - after rarefied). ANCOM too produced another output with the information on either null hypothesis is rejected or not. These three taxa were selected from the list. Are these taxa contributing to the separation of microbial communities for at least two treatments? Am I correct? This has puzzled me for some time.

Thank you once again.

Hello @Yit_Goh,
I was wrong in my previous message about ANCOM. I misunderstood how Ancom works. The W value is a measure of how many of your ratio pairs hit the set significance threshold value. I believe that QIIME2 sets this critical value to .05. Therefore I was misinformed in my statement that you can not claim significant differences. You most definitely can!

You learn something new every day!

However, I am wondering why you rarefied your data before ANCOM? ANCOM is self-normalizing in regards to sequencing depth so a rarefied table is unnecessary in most situations.

Yes. They are contributing to the differences in the microbial community. The don’t account for all of the differences but ANCOM is pointing out species that are abundantly different in your groups which shows one of the differences between your two groups. Does that answer your question?

I hope this helps!
Chloe

3 Likes

Hi @cherman2,

Thank you so much. This is really helpful.
My apologies. The data was filtered prior to ANCOM.
Yes, you did answer the question on the contributing microbial taxa.

Thank you once again.

1 Like