ANCOM clr results confusion

ancom
(Zach Burcham) #1

Hello,

I am having some confusion with my ANCOM results. From my understanding, the clr is like the fold-change and the 100% abundance shows the counts in each group. I would expect if I have group A and B then anytime I have more counts in A than B I would obtain either positive or negative clr’s consistently. But in my data set, I have two significant taxa that are both higher in group A than B, but one of them gives a negative clr while the other gives a positive clr. How is this possible? Do I report these taxa as being more abundant in group A based on the percentile abundance or do I report one as more abundant and the other less since the clr’s are positive and negative even though the percentile abundace says both are more abundant?

Thanks,
Zach

0 Likes

(Nicholas Bokulich) assigned mortonjt #2
0 Likes

(Jamie Morton) #3

Hi @Zach_Burcham - having a plot showing the results could help us better answer this question.

clr coordinate do not necessarily imply increase / decrease (since it is not possible do this in the first place). Instead, these coordinates represent log-fold change relative to the average microbe.

So it is possible that both taxa are higher, but one of them is increased greater than the average microbe and the other is increased less than the average microbe (provided that the average microbe is also increased between these two conditions).

1 Like

(Zach Burcham) #4

ancom-adult_or_youth-L6.qzv (465.9 KB)

Hi @mortonjt - Thank you for your response. I have attached a .qzv file from my data to better show what I mean. For example, Treponema and Lautropia both show higher counts in the Adult samples and the CLR is practically opposite. So from my understanding of the “average” microbe approach, CLR isn’t a way to determine which sample group contained higher relative abundances. So is it appropriate to say that the relative abundances of Treponema and Lautropia are higher in Adult samples than Youth when reporting these results?

0 Likes

(Jamie Morton) #5

Correct - no one can actually do differential abundance without absolute abundances: see below (the clr transform is also described there)

https://bmcgenomics.biomedcentral.com/articles/10.1186/s12864-018-5160-5

You can report Treponema and Lautropia have differing relative abundances according to ANCOM. I wouldn’t put too much weight on the clr transform in the plot – that is just to help visualize the effect size.

3 Likes

(Zach Burcham) #6

Great, thanks for the help!

0 Likes

(Nicholas Bokulich) unassigned mortonjt #7
0 Likes

(Zach Burcham) #8

Hi @mortonjt, one more question. When looking at my feature table, it shows higher counts of Neisseria in my youth group samples compared to adults, but when I run ANCOM I get that there is a significant difference between my age groups with Neisseria being higher in my adult samples. Why does there seem to be discrepancy here? The other genera ANCOM is telling me are significantly different correspond to the feature table except one.

0 Likes

(Matthew Ryan Dillon) assigned mortonjt #9
0 Likes

(Jamie Morton) #10

Really not crazy about using the terms lower or higher in the context of relative abundances, since relative abundances can’t tell you which microbes actually increased or decreased.

I’m not an expert on ANCOM - best consult with the corresponding author on the ANCOM paper

0 Likes

(Nicholas Bokulich) unassigned mortonjt #11
0 Likes