Mrbayes Produces Incorrect Tree When Run In Mpi Mode

Question

[Pappu]

I constructed a phylogenetic trees of a set of 4 distantly related protein families. In FastTree WAG+CAT20, two protein families seem to come from a common ancestor and the other two families for separate clusters. I used the same alignment in MrBayes 3.2.2 MPI compiled with Beagle library with WAG+I+G (2 runs, 1 million generations each, sampling every 100th run) which makes a separate nodes for most of the sequences:

|----a 
|
|----b
|
|----c

I am wondering what went wrong. The alignment seems to be all right. It seems like something is wrong with the compiled version with MPI library which is quite fast. Single threaded version can make branches properly.

Answer 1

[neal.platt]

We have the same problem (large polytomies). Take your data and run the "sumt" command in MrBayes3.1 and you will recover the appropriate tree.