ANCOM volcano plot - high W negative clr

Hi qiime2 community, I have a volcano plot interpretation question resulting from my ANCOM analysis

Having ran ANCOM at the genus (level 6) and species level (7) comparing a control group to a lesion group . and it gave me interesting results: Treponema species has a high W – so they are standout taxa that qiime pointed out.

These really high W-statistic numbers have negative clr – and I am unsure why that is. Does it actually mean it is decreasing in abundance across the categories? I was under the impression that usually taxa on the upper right are the ones that are differentially abundant where taxa increased abundance.

And included in the output is the abundance table – and it is pretty clear that these species are clearly more abundant in the lesions than in the control samples. So WHY are these taxa despite the high W’s have a NEGATIVE clr.

I have also other examples at different taxonomic levels and found taxa (genus, family, etc) have high Ws and positive clr’s.

How do I properly interpret this – Should I basically disregard whether it’s on the left or the right side of the graph (+/- clr) because the W number is so high?

Thanks a bunch, and appreciate any feedback and for reading this,

Nick

Hi @mobynick ,

Welcome to the :qiime2: forum!

In your volcano plot, the X-axis (i.e. the clr axis) represents the effect, or the impact that each individual feature has with respect to the rest of the community within your samples. The Y-axis is the distribution of your W scores. The values that have the highest W scores are likely the most differentially abundant (or the most important) in your sample.

The effect (your X-axis) is essentially just a way to demonstrate the distribution of your samples - and the positive/negative values will change depending on if you are using the centered log-ratio of this variable, or a different representation.

With that being said, the positive or negative values of the clr (which stands for centered log-ratio) of the effect aren't something I would get hung up on here. It's the W score that is really relevant here, which as you mentioned is associated with the two species you see in your abundance table that are the highest.

For more details on how to interpret these volcano plots (along with a couple of helpful examples), I would highly recommend checking out this QIIME 2 tutorial video on ANCOM Differential Abundance Testing. This should help you better understand what you are looking at in these distributions!

Hope this helps!

Cheers :qiime2: :lizard:

3 Likes

That helps! thank you for the reassurance. But is there a particular reason why this is the case? what does it mean to have negative clr as opposed to positive clr, and the view that we should look for features in 'top right corner" of plot

Is it recommended to try to tinker with the ancom parameters - ie
--p-transform-function TEXT Choices('sqrt', 'log', 'clr')
or
--p-difference-function TEXT Choices('mean_difference', 'f_statistic')

What would these options entail as opposed to the defaults?

Thanks so much again,

Nick

Hi @mobynick,

For the effect axis (which we are viewing as the centered log-ratio, or the clr), the positive vs. negative is telling you in which direction the feature is affecting the rest of the community within your samples (whether you are seeing more of Feature A vs. Feature B, C, etc or vice versa).

I'm not sure where the suggestion of only looking in the top right corner of these plots came from - but that is not a good rule of thumb. I can't speak for your specific research project, but when analyzing differential abundance, it is typically beneficial to be looking at the features with the highest differential abundance (or W score) - which will be present at the top of the Y axis (but not necessarily in just the upper right hand corner). These are ANCOM differential abundance plots are typically called volcano plots because you'll see features with a high W score on either corners of the plot (thus creating a volcano or crater-like shape).

All of that being said, I would highly recommend watching that tutorial video I linked above. This video discusses all of this (and more!) in great detail, and can further clarify all of the questions you've asked above.

Cheers :lizard:

3 Likes