Further clinical studies are needed to confirm whether treatment of all nerves branches which cross the treatment line is required to provide a meaningful clinical benefit for patients or if treatment of a subset of branches is sufficient.
The potential existence of rare positively selected sites in a subset of branches was then assessed.
The branch-site model (BSM) [ Yang, 2006 (Section 8.4); Zhang et al., 2005] allows to detect positive selection that affects a subset of codon sites for a subset of branches in a phylogenetic tree.
Their model notably differs from the BSM (Zhang et al., 2005) in that putative positive selection is not optimized on a priori defined branches, but over a subset of branches which are determined by the software.
Since the subset of branches to be tested has to be specified a priori and independently of the results of the branch-specific test, we decided to focus on those branches defining the major taxonomic groups of the vertebrate phylogeny, numbered from B1 to B13.
A model in which ω was fixed across the tree (one-ratio) was compared with models in which ω was allowed to differ in a subset of branches (two- and three-ratio models), and the significance of the difference was assessed using the likelihood ratio tests (LRTs).
As expected, the Brownian motion model exhibits an "averaging effect" more generally, in which transformations in body mass are distributed somewhat evenly over the phylogeny, while the stable model permits large transformations in body mass to occur on a smaller subset of branches.
We compared the inference restricted with specific subsets of branches.
CODEML was then used to obtain maximum likelihood estimates of the dN/dS ratio per site for the whole tree (basic model [ 46]) or independently for subsets of branches (e.g. trunk vs. non-trunk branches, branch model [ 47].
RELAX tests whether selection is relaxed or intensified on a subset of test branches compared with a subset of reference branches in a predefined tree.
