So, if you use the fixes from the posts above, you should be able to output a minimum change in each base from the transcription factor binding site. In the top up phase it looks like you are clustering the individual bases to determine the sequence variability and relations in the sequence itself. By recursively clustering these, you should be able to output a second value that serves as a "score" for the sequence variation, which then allows you to cluster again in the top down version to determine the genotypic relation of the various species. A simple way to cluster them would be to use a sub to recursively find the difference of the "values" your getting for each base associated with each species, pushing them into an array, and them using them to create a second value which is then used to determine the larger phylogenetic tree. It would be more efficient to use a matrix to sort and output those values so that the changes would be more fluid and dynamic and use less memory, etc. Bioperl has a couple modules that are already made to determine the distances in the various nodes, so you could integrate these into your program to get the larger phylogenetic output you desire.

Check out Module:Bio::Tree::Node, as well as Module:Bio::Align::DNAStatistics, which if I recall can do something of what your trying to do above. Good luck!