The thumbnail sketch of the methodology for detecting covariation that is presented on datamonkey.org says that the method finds covariation between sites that change on the same branch of a phylogeny.  However, from my first reading of Poon et al (2007) (PLoS CompBiol: An evolutionary-network model ... ) it was not obvious to me that the method was limited to covariation between sites that changed on the same branch.  Am I missing something in Poon et al?  What is the true story?

Hi Catherine,

The power to detect covariation is (loosely) based on the number of branches where two sites have a substitution, i.e. if two sites covaried perfectly, you would see that they always experience concurrent substitutions. In Poon 2007 there was another method about the covariation of RATES, not substitutions -- that is limited to pairs of sites, however, and not implemented in Datamonkey.

Sergei

Hi Sergei:

I think I have answered this question by realizing that unless you have a meaningful of ordering branches, it won't be possible to pick up a signal from covariation such as when a change at one site is permissive for a subsequent change at another site.

The Poon article to which I referred is the one in which you infer a Bayesian network for HIV V-3 , and uses ancestral reconstruction etc.  I don't think we are speaking of the same articles.

Catherine

Hi Catherine,

Right: the paper I was referring to is Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login

Sergei