Dear Andrew,

Firstly, I would encourage you to look over the book chapter on selection using HyPhy that we recently wrote (http://www.hyphy.org/pubs/hyphybook2007.pdf) to see what HyPhy does and how it is similar/different from PAML. HyPhy includes a couple of analyses which could run exactly the same models as PAML, but also a much larger set of tools which is methodologically different from PAML (this is also discussed in the chapter). We also have an online server for some of the selection analyses at www.datamonkey.org. A long time ago I verified that when the same models are used (e.g. M8), HyPhy and PAML do return the same results (within a few decimal places anyway).

Some of the analyses are designed to be very fast (e.g. SLAC), and some are rather intensive (e.g. those which use genetic algorithms).

HyPhy is multithreaded and will try to take advantage of all available cores in an SMP system. On sufficiently large codon datasets I am able to load all 8 cores in my dual-quad Mac system. HyPhy can also take advantage of an MPI cluster for a large number of analyses - especially those which do the same task repeatedly (simulations or site-by-site fitting would be good examples).

HTH,
Sergei