Interpretation of differences in PD wholetree

[Rune Grønseth]

Hi, I'm sorry if there has been similar posts previously, but I haven't been able to find them.

I've compared 6 different sampling methods (on the same subjects) and get the following mean pd_wholetree metrics:

group 1: 35.9

group 2: 30.6

group 3: 29.9

group 4: 27.6

group 5: 26.2

group 6: 22.2

All comparisons between groups (paired t-tests and non-parametric paired signed rank test) are highly significant as well as a non-parametric test for an ordered trend. 

However, could somebody tell me how to know wether these rather small numerical differences (at least for groups 2-5) bear any biological meaning? 

Cheers,

 Rune 

[Colin Brislawn]

Hello Rune,

The biological meaning of the different numbers stems from the alpha diversity metric you are using. So another way to ask this question is to say 'what does the pd_wholetree metric actually measure?'

This metric name stands for Phylogenetic Diversity, whole tree, and was published here. 

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2674678/ 

I hope that helps,

Colin

[Rune Grønseth]

Thank you, Colin.

I was thinking more in terms of how the actual numerical difference can be interpreted? Would a difference between group 2 and 3 bear also be biologically important? In medicine we often talk about minimal clinically important difference as we often find statistically significant differences that don't bear clinical meaning?

Did that make sense? 

Best wishes from Rune

[Colin Brislawn]

Hello Rune,

I understand the distinction you are making about significance; just because something is statistically significant, does not mean that the magnitude of difference makes it's biologically significant.

Would a difference between group 2 and 3 bear also be biologically important?

I'm not sure... Can you help me understand the biological question you are pursuing here? Perhaps an alpha diversity metric like Faith's PD is not the best way to answer your question. I often find that differential abundance testing or beta diversity metrics are more useful to address my questions.

Thanks for telling me more,
Colin
 

[Rune Grønseth]

Thanks Colin, 

We have sampled the lungs using oral (group1), a small-volum "lung wash (lavage)", different fractions of protected lung wash(bronchioalveolar lavage, groups 3 and 4), protected brushes (group 5 and 6). We're looking at beta-diversity and differential abundance testing (Deseq2) as well, and the differences in alpha-diversity is kind of consistent with what we would expect, but we are unsure on how to interpret (if at all) the numerical differences between the groups.

Cheers,

Rune

[Colin Brislawn]

Good morning Rune,

Ah, so using this alpha metric, you could should that the oral sampling (group1) captures more microbes (increased richness) compared to the lavage (group2). Maybe this is a biological interesting finding, or perhaps this informes your study design or interpretation. Differential abundance testing would be perfect to test for microbes which are more common in group 1. I think you are on the right track.

I'm afraid I may not be much help here. I'll reach out to a qiime dev who may know more about this metric. 

Colin