For positive selection prediction (Codeml module of PAML), one needs to provide the species tree. The Ids of the tree should be same as in the alignment file. Alignment file can be generated by codon based alignments (prank-codon is the recommended package).
I'm a little confused by your question. PAML is a maximum likelihood phylogenetics program which can handle both nucleotide and protein data. The input would be an aligned data matrix of nucleotide or protein data and the output would be a nexus file corresponding to the phylogenetic tree (or trees). There should be no input tree, but perhaps there is an analysis in PAML that I am not aware of. It's not on my list of choices for best phylogenetics programs.