convergence issues

30 views
Skip to first unread message

Matthew Prebus

unread,
Sep 30, 2019, 3:17:30 PM9/30/19
to ExaBayes
I'm attempting to run a 25 taxa, 193K bp dataset with 34 partitions, but I have been having issues with convergence in several parameters between runs, even after 10 million generations. I've set up the configuration file as follows:

#NEXUS

begin params;
  brlens = (0-33)
end; 

begin proposals ;
  eTBR         5
  eSPR         0
  stNNI        0
  parsimonySPR 5
  likeSpr      5
 
  branchMulti    15
  treeLengthMult 1
  nodeSlider     5
  blDistGamma    0
end;

begin run;
  numruns 4
  numGen 10000000
  numswapspergen 4
end;

Any help would be appreciated.

Alexandros Stamatakis

unread,
Oct 16, 2019, 5:45:04 AM10/16/19
to exab...@googlegroups.com
Hi Matthew,

Sorry for the late reply, I was on parental leave.

It may well be that your dataset does not converge as it exhibits two or
more areas of high posterior probability.

If I remember well, a particularly tricky one is this dataset with 27
taxa reported by Lakner at all in here:

https://academic.oup.com/sysbio/article/57/1/86/1704335

You can find the dataset here:

https://github.com/stamatak/BioinfProtocols

as I stumbled upon it by chance when writing this book chapter

https://currentprotocols.onlinelibrary.wiley.com/doi/abs/10.1002/0471250953.bi0614s51

and rediscovered the problem with the two likelihood peaks under ML.

So my suggestion is that you try re-analyzing the dataset with ML and
see if you get several peaks in the likelihood surface there as well to
be more confident that the dataset is difficult.

Hope this helps,

Alexis
> --
> You received this message because you are subscribed to the Google
> Groups "ExaBayes" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to exabayes+u...@googlegroups.com
> <mailto:exabayes+u...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/exabayes/3b809fb1-839a-4fa1-86bb-3f894207c08a%40googlegroups.com
> <https://groups.google.com/d/msgid/exabayes/3b809fb1-839a-4fa1-86bb-3f894207c08a%40googlegroups.com?utm_medium=email&utm_source=footer>.

--
Alexandros (Alexis) Stamatakis

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

www.exelixis-lab.org
Reply all
Reply to author
Forward
0 new messages