If dN/ds ratios are allowed to “vary among branches” (a common phrase in the literature), what does that acutally mean? It seems that dN/ds ratios can only be calculated when two or more sequences are being compared. Let’s say you have 4 sequences, each representing the tip of a tree. Branches on this tree would represent the divergence in the composition of those 4 sequences, and the length of the branches would be proportional to the level of divergence. So if the dN/ds ratio is allowed to vary for each branch, does that mean that a separate ratio is calculated for each pairwise comparison?