I'm struggling to understand application of branch, site and branch-site models to detect selection pressure across nucleotide coding sequences (CDS) of the set of genomes i'm working with. After reading of PAML manual and several codeml tutorials certain things are still not completely clear for me. Could you clarify these for me please?

1. Are the site models only used to detect selection pressure in certain genes of a genome? I.e. there are 6 alignments for different sets of orthologous genes and i'd like to know which alignments (genes) are under selection pressure and which are not. The branch and branch-site models could not be applied for that. Am i right?

2. Why we need a phylogenetic tree even for application of the site models ? How tree topology will influence on the results of the analysis by means of site models?

3. Are the branch and branch-site models typically applied for detecting differences in selection pressure between branches of the phylogenetic tree? If so, which the most important differences in usage between branch and branch-site models? I mean the typical use cases for both model classes.

4. Are the clade model C and clade model D represent a two separate model classes or just a subclasses of branch-site model? What is the typical application for the clade model C and clade model D?

Thanks,

Denis