Estimating accurate Gamma-based branch lengths for large phylogenies with RAxML

[Jean-Baka Domelevo Entfellner]

Hi there,

We would like to know if it is reasonable to ask RAxML to estimate accurate branch lengths on a phylogeny with 12,000+ taxa with RAxML? Giving a fixed topology and asking RAxML to optimize branch lengths using a standard Gamma law for substitution rate heterogeneity across sites, would that be successful? No overflow/underflow numerical errors to foresee? No automatic switch to another rate heterogeneity model?

Thanks for your answer,
   JB

[Alexey Kozlov]

Hello Jean-Baka,

a short answer would be: you should try :)

If it's a single-gene alignment, branch length/model optimization (-f e) should be pretty fast.

As long as you specify "-m GTRGAMMA", no automatic model switching will occur.

As for your other concern, numerical underflow problems might really happen with the dataset of this size - it depends
on your alignment/topology. We do have a solution for this, but unfortunately it's not (yet) integrated in the
production version of RAxML. So should standard RAxML fail, you can try the following branch which contains the fix:

https://github.com/amkozlov/raxml-sativa

We would also like to hear about the outcome, to have an idea how often the problem surfaces in practice. It will allow
us to set reasonable defaults in the future versions of RAxML, since preventing numerical underflow incurs some overhead.

Best,
Alexey

> --
> 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 Stamatakis]

Thanks Alexey :-)

Also, the underlying numerical problem with GAMMA is discussed in this
paper here:

https://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-12-470

Alexis

--
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

[Bui Quang Minh]

Dear Alexis and Alexey, 

I am quite interested in the technical details how you solved this numerical issue... can you share it?

Thanks ;-)

Minh

To unsubscribe from this group and stop receiving emails from it, send an email to raxml+un...@googlegroups.com.