Multigene phylogeny

[Antony03]

Hi!
I would like to do a phylogeny based on multigene alignment. I have 97 differents genes, I know how to align these genes (with prank for example). If I make different tree (one per gene) with Mrbayes, is the command ''sumt'' will combine all tree in one?

Thanks and if you have any suggestion, feel free!

[A_Morozov]

Well, sumt them doesn't sound really great. You should either make a tree for each of them and then find consensus tree, or concatenate alignments and define different models for them (ie allow each alignment's evolutionary process to be estimated independently but look for single tree for them all; look at manual). Latter is easier to do, but first seems better because it allows you to check manually whether topologies are significantly different and whether there were any putative horizontal transfer events.

[kmkocot]

Antony,

Running sumt on many single-gene trees will give you a supertree, I suppose, but that's not what is generally done.

Check out the programs FASconCAT and SCaFoS. They let you concatenate multiple alignments into one supermatrix. If you are interested in using maximum likelihood in RAxML, I can give you a script that will give you the best-fitting model for each partition and print out a raxml partition file.

Best,
Kevin

[A_Morozov]

Also check out "beast", which is another bayesian inference tool and sports a nice GUI that allows to define models and partitions and such without knowing programming and file format specifics.

[ssing]

You might want to consider BUCKy: http://www.stat.wisc.edu/~ane/bucky/index.html. This program summarizes across individual gene trees to show the most common topology for the species tree.