Suprising results from acestral state reconstruciton

[jodyp...@gmail.com]

Hi,

I would like to perform some ancestral state reconstruction for a large dataset. The alignment consists only of SNP sites.

I ran some tests on a small number of the taxa.
After creating a tree and rooting it I run:
raxmlHPC-PTHREADS-SSE3 -f A -m GTRGAMMA -n test -p 12345 -s snps.fasta -t rooted.tree --asc-corr=lewis

Here I have shown the states superimposed on the tree for one site:

       /-H37Rv1, G
      |
      |              /-por6B, T
      |         /11, T
-ROOT, A       |     \-por6A1, T
      |    /10, T
      |   |    |     /-por5B, T
      |   |     \12, T
       \9, G         \-por5A1, T
          |
          |    /-por7A1, G
           \8, G
               \-por7B, G

To me it looked a bit strange that an "A" has been assigned to the root.
The probabilities are: 0.250328 0.249678 0.250303 0.249691
I have little understanding on the underlying model and how it reaches these probability, but just by intuition it seems like it should be a G.

Is there anything wrong with the method I am using? or is there really an (almost) equal probability for each state?

Thanks,
Jody

[Alexandros Stamatakis]

This does not look strange, as the probabilities are almost identical
and this is not parsimony, that is, the ancestral state depends on the
branch lengths, the stationary frequencies, and the substitution model
parameters.

What I am not so sure about is if ancestral state reconstruction works
correctly with the ascertainment bias correction.

You could try executing this without asc bias correction just to confirm,

alexis

> --
> You received this message because you are subscribed to the Google
> Groups "raxml" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to raxml+un...@googlegroups.com
> <mailto:raxml+un...@googlegroups.com>.
> For more options, visit https://groups.google.com/d/optout.

--
Alexandros (Alexis) Stamatakis

Research Group Leader, Heidelberg Institute for Theoretical Studies
Full Professor, Dept. of Informatics, Karlsruhe Institute of Technology

www.exelixis-lab.org

[jodyp...@gmail.com]

Thanks I will try without asc bias correction.