bray curtis vs unifrac

[Jo87]

Hi,

If I observe a significant difference between two groups of samples when using the bray-curtis metric, but not when using the unifrac metric (beta-diversity analysis), is it correct to say that the OTUs which differ between the two groups are phylogenetically close to each other, and therefore there is a high level of shared branch length in the community which would not generate a significant difference when using the unifrac distance(s)?

If the changes between the two groups were based on OTUs not sharing the same branches (more phylogenetically distinct), then we would see a significant difference when the unifrac metric(s) were applied?

Is this understanding correct? 

Thank you for your help,

Jo

[Jessica Metcalf]

Hi Jo, 

Yes, what you describe is a likely explanation for your results. You can always confirm by taking a look at the taxonomy plots. It is also worth looking at weighted unifrac, which takes into account relative abundance of OTUs as well.  

Thanks,

Jess

[Colin Brislawn]

Wow.

"OTUs which differ between the two groups are phylogenetically close to each other, and therefore there is a high level of shared branch length in the community which would not generate a significant difference when using the unifrac distance"

That's a great description the UniFrac metric.

Well said.

Colin