Differences between binary and multistate mode for binary data

[Shunhua Han]

Hi,

I tried to use a binary matrix to generate a tree, but I couldn't figure out which substitution model (MK/GTR/unorder) is used when in BINCAT/BINGAMMA  from the manual. I suspect BINGAMMA is exactly the same as MULTIGAMMA + GTR model when I use binary data, and the experiment using both methods gave similar tree. Could anyone confirm that? Appreciate it!

[Alexandros Stamatakis]

none of the MK/GTR/UNORDERED is used, there's only one relative rate in
the binary data substitution model, so essentially there is only one
time-reversible model we can come up with for binary data ....

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
Adjunct Professor, Dept. of Ecology and Evolutionary Biology, University
of Arizona at Tucson

www.exelixis-lab.org

[Shunhua Han]

Thanks for you reply!

That raxml is not using GTR-like model in binary mode surprises me. But If I understand what you say correctly, the state frequencies for 0 and 1 are plugged in the binary data substitution model, which is something like GTR (as showed in the graph below, a=1, 𝜋1, 𝜋0 is the equilibrium distribution and 𝜋1+𝜋0=1),  And I guess that's the reason I got trees from MULTIGAMMA and BINGAMMA having basically same topology and branch length. If not, could you point out the detailed difference between those two methods?

Also, I am wondering if ascertainment bias correction would make a difference to the tree topology, or in a more detailed way, whether raxml adds certain amount of invariant sites to the matrix, or just changes the branch length values based on some algorithm after best tree is calculated?

Thanks for your help!

[Alexandros Stamatakis]

On 15.05.2017 22:31, Shunhua Han wrote:
> Thanks for you reply!
> That raxml is not using GTR-like model in binary mode surprises me.

it is using a time reversible model, only that for two states there is
only one single such model!

> But
> If I understand what you say correctly, the state frequencies for 0 and
> 1 are plugged in the binary data substitution model, which is something
> like GTR (as showed in the graph below, a=1, 𝜋1, 𝜋0 is the equilibrium
> distribution and 𝜋1+𝜋0=1),

yes that's correct.

> And I guess that's the reason I got trees
> from MULTIGAMMA and BINGAMMA having basically same topology and branch
> length. If not, could you point out the detailed difference between
> those two methods?
>
>

> Also, I am wondering if ascertainment bias correction would make a
> difference to the tree topology, or in a more detailed way, whether
> raxml adds certain amount of invariant sites to the matrix, or just
> changes the branch length values based on some algorithm after best tree
> is calculated?

see here:
https://academic.oup.com/sysbio/article/64/6/1032/1669226/Short-Tree-Long-Tree-Right-Tree-Wrong-Tree-New
for an answer,

alexis

>
> Thanks for your help!
>
>
> On Monday, May 15, 2017 at 11:18:14 AM UTC-4, Alexis wrote:
>
> none of the MK/GTR/UNORDERED is used, there's only one relative rate in
> the binary data substitution model, so essentially there is only one
> time-reversible model we can come up with for binary data ....
>
> alexis
>
> On 15.05.2017 02:29, Shunhua Han wrote:
> > Hi,
> >
> > I tried to use a binary matrix to generate a tree, but I couldn't
> figure
> > out which substitution model (MK/GTR/unorder) is used when in
> > BINCAT/BINGAMMA from the manual. I suspect BINGAMMA is exactly
> the same
> > as MULTIGAMMA + GTR model when I use binary data, and the experiment
> > using both methods gave similar tree. Could anyone confirm that?
> > Appreciate it!
> >
> > --
> > 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 <javascript:>
> > <mailto:raxml+un...@googlegroups.com <javascript:>>.

> > For more options, visit https://groups.google.com/d/optout

> <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
> Adjunct Professor, Dept. of Ecology and Evolutionary Biology,
> University
> of Arizona at Tucson
>

> www.exelixis-lab.org <http://www.exelixis-lab.org>

[Shunhua Han]

Thanks for the paper you recommended, that definitely clears my confusion! :)

Shunhua