Generalized UniFrac question

Hi Everyone,

I’m using the beta-phylogenetic-alt plugin to calculate generalized UniFrac distances. There is an option in this plugin to perform variance adjustment, what I did, and the command completed successfully.

My question is: can I do this using this distance matrix, or this option in only valid for weighted unifrac (based on the previous paper: Chang et al. BMC bioinformatics, 12(1):118, 2011).

Thanks very much in advance,
FS

1 Like

Hello Fernando,

OK great!

Are you asking if --p-variance-adjusted True is the same as the method proposed in Chang (2011), because yes that’s the same method.

I’m not sure I understand your question.

Are you asking if there is option for ‘Variance Adjusted Unweighted UniFrac’? The Variance Adjusted method adjusts the way weights are used, and in Unweighted UniFrac the weights are not used at all… so no there is no option for Variance Adjusted Unweighted UniFrac.

Let me know if this is helpful,
Colin

Hi Dr. Colin,

Thanks for your reply.

What I wanna ask is: can I use a Variance Adjusted Generalized UniFrac? This distance, from what I understood, also takes into account weights, as it combines both weighted and unweighted versions. Or the variance option is only correct when using weighted UniFrac.

Thanks,
FS

2 Likes

Now I understand your question!

That’s an interesting idea. I’m not sure!

Let’s see what @Nicholas_Bokulich recomends.
Colin

@fstudart, it is accurate that Chang et al. only defined the variance adjustment over weighted UniFrac. However, the modification to the formula can be readily extended to other UniFrac variants (including unweighted, and generalized), and we did not see a practical reason to not allow this operation.

Best,
Daniel

2 Likes

To clarify, with Striped UniFrac (beta-phylogenetic-alt), you can apply variance adjustment to unweighted, weighted normalized, weighted unnormalized, or generalized UniFrac. To the best of my knowledge, the adjustment has only been explored in the context of weighted normalized UniFrac.

Generalized UniFrac also does not exactly combine weighted and unweighted; an alpha of 1.0 is identical to weighted normalized UniFrac, but an alpha of 0.0 is only approximately unweighted UniFrac as it does not dichotomize the proportions.

Hope that helps,
Daniel

2 Likes

Hi Daniel,

Yes, It definitely helps.

Thanks very much for your feedback,
FS

1 Like

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.