Hello Evoldir. Here are the replies to my request for programs that could do ML phylogenetic reconstruction with DNA data partitioning, such as when one has multiple genes and wants to run by codon position by gene partitions (as one can do in MrBayes). exact response listed below. My summary: TREEFINDER (www.treefinder.de) or RAxML (http://icwww.epfl.ch/~stamatak/index-Dateien/Page443.htm) or HyPhy www.hyphy.org thanks for all the feedback and help, chris 1) Hi, Chris. I've created such a program as part of my PhD dissertation, which I'll be completing this summer. The manuscript is in prep; I won't distribute the program until its approach has been peer-reviewed and the paper is at least in press. Preliminary comparisons, running the same non-partitioned models in PAUP and in the new program, suggest that it can be up to 20 times faster than PAUP at finding the ML topology, plus the program can run partitioned models, too (but I haven't evaluated its speed for that, yet). One thing you could try in the interim (and assuming that you don't find other programs which can already do partitioned ML quickly) is to do site-specific rate models in PAUP. It won't take into account different base frequencies or transition matrices across partitions, but will let you allow different partitions to at least have different average rates. Also, five years ago, it was announced that there would be a "bleeding-edge" version of PAUP which would have new goodies such as partitioned models: I haven't heard anything about this since then, but you could check with the developers. 2) To my knowledge, the only program (other than codeml) that can do this currently is RAxML. You can download it at: http://icwww.epfl.ch/~stamatak/index-Dateien/Page443.htm I believe that Derrick Zwickl is also working on implementing partitioned models in Garli, although I don't think it's ready to go yet. You can check for Garli updates at: http://www.zo.utexas.edu/faculty/antisense/garli/Garli.html 3) TREEFINDER (www.treefinder.de) is a ML program that will accept multiple gene partitions along with different rate heterogeneity models. It will assume that branch lengths are proportional between gene partitions. RAxML (http://icwww.epfl.ch/~stamatak/index-Dateien/Page443.htm) is a very fast and efficient ML program that will allow for the existence of different GTR(Gamma or CAT) models in a multipartition analysis. It estimates a single branch length across partitions. By the way, codeml is not designed for phylogenetic reconstruction. 4) Our open-source phylogenetic software package, called HyPhy, is particularly well-suited for your problem. Please feel free to check it out at www.hyphy.org 5) If you are working with RNA genes, then try PHASE. Christopher W. Wheat, PhD Postdoctoral Researcher Pennsylvania State University University of Helsinki University of Miami http://www.helsinki.fi/science/metapop/english/People/Chris.htm Christopher West Wheat