--bootstop-perms & -# autoMRE
27 views
Skip to first unread message
James Munro
Aug 6, 2021, 11:02:51 AM8/6/21
to raxml
Hi,
I have a dataset that will not reach convergence for the 1,000 BS reps called on using autoMRE. I have tried two approaches to try and figure out how many BS reps might be needed for my "final" analysis.
1) using -I and the a posteriori bootstopping analysis:
I ran 3 independent analyses with 5,000 BS reps, concatenated the tree files and ran $ raxmlHPC-PTHREADS-AVX -m GTRCAT -z algn.bs -I autoMRE -p 12345 -T 2 -n TEST. This was successful and let me know that just under 5,000 reps are needed.
2) using --bootstop-perms and autoMRE:
This has me puzzled. I have read through the output files, yet I see no suggestion as to convergence as one would typically see when calling autoMRE, such as "Stopped Rapid BS search after 250 replicates with MRE-based Bootstopping criterion". I ran $ raxmlHPC-PTHREADS-AVX B 0.03 --bootstop-perms=1000 -b 12345 -p 12345 -# autoMRE -s algn.fasta -m GTRCAT -n TEST.
I have two questions:
Am I on the right track with the a posteriori analysis?
Should --bootstop-perms and autoMRE be outputting a suggested number of suitable BS reps or am I misunderstanding the intention of this?
Thanks!
James
Alexandros Stamatakis
Aug 25, 2021, 11:47:46 PM8/25/21
to ra...@googlegroups.com
Hi James,
> I have a dataset that will not reach convergence for the 1,000 BS reps
> called on using autoMRE. I have tried two approaches to try and figure
> out how many BS reps might be needed for my "final" analysis.
>
> 1) using -I and the a posteriori bootstopping analysis:
> I ran 3 independent analyses with 5,000 BS reps, concatenated the tree
> files and ran $ raxmlHPC-PTHREADS-AVX -m GTRCAT -z algn.bs -I autoMRE -p
> 12345 -T 2 -n TEST. This was successful and let me know that just under
> 5,000 reps are needed.
Okay, but this is a very high number and it might indicate that there is
little signal in your data or maybe that there exist multiple local ML
optima for your dataset. Please have a look at:
https://currentprotocols.onlinelibrary.wiley.com/doi/abs/10.1002/0471250953.bi0614s51
where I describe such a case (non convergence after 1000 reps) for a
very small dataset of 27 taxa that is known to be difficult to analyze
as it has two local ML optima.
> 2) using --bootstop-perms and autoMRE:
> This has me puzzled. I have read through the output files, yet I see no
> suggestion as to convergence as one would typically see when calling
> autoMRE, such as "Stopped Rapid BS search after 250 replicates with
> MRE-based Bootstopping criterion". I ran $ raxmlHPC-PTHREADS-AVX B
> 0.03 --bootstop-perms=1000 -b 12345 -p 12345 -# autoMRE -s
> algn.fasta -m GTRCAT -n TEST.
>
> I have two questions:
> Am I on the right track with the a posteriori analysis?
Yes, I'd say so, but I would have a closer look at the ML trees you
inferred on the original MSA if I was doing this analysis.
> Should --bootstop-perms and autoMRE be outputting a suggested number
> of suitable BS reps or am I misunderstanding the intention of this?
The maximum number of BS reps under autoMRE is 1000, the
--bootstop-perms setting is an internal setting for the actual
convergence test that is being performed, and does not set the number of
BS replicates autoMRE will execute before giving up.
Alexis
>
> Thanks!
>
> James
>
> --
> 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>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/raxml/1edc545a-470c-4cb5-b2ec-851ce5c6ce71n%40googlegroups.com
> <https://groups.google.com/d/msgid/raxml/1edc545a-470c-4cb5-b2ec-851ce5c6ce71n%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
Affiliated Scientist, Evolutionary Genetics and Paleogenomics (EGP) lab,
Institute of Molecular Biology and Biotechnology, Foundation for
Research and Technology Hellas
www.exelixis-lab.org
> I have a dataset that will not reach convergence for the 1,000 BS reps
> called on using autoMRE. I have tried two approaches to try and figure
> out how many BS reps might be needed for my "final" analysis.
>
> 1) using -I and the a posteriori bootstopping analysis:
> I ran 3 independent analyses with 5,000 BS reps, concatenated the tree
> files and ran $ raxmlHPC-PTHREADS-AVX -m GTRCAT -z algn.bs -I autoMRE -p
> 12345 -T 2 -n TEST. This was successful and let me know that just under
> 5,000 reps are needed.
little signal in your data or maybe that there exist multiple local ML
optima for your dataset. Please have a look at:
https://currentprotocols.onlinelibrary.wiley.com/doi/abs/10.1002/0471250953.bi0614s51
where I describe such a case (non convergence after 1000 reps) for a
very small dataset of 27 taxa that is known to be difficult to analyze
as it has two local ML optima.
> 2) using --bootstop-perms and autoMRE:
> This has me puzzled. I have read through the output files, yet I see no
> suggestion as to convergence as one would typically see when calling
> autoMRE, such as "Stopped Rapid BS search after 250 replicates with
> MRE-based Bootstopping criterion". I ran $ raxmlHPC-PTHREADS-AVX B
> 0.03 --bootstop-perms=1000 -b 12345 -p 12345 -# autoMRE -s
> algn.fasta -m GTRCAT -n TEST.
>
> I have two questions:
> Am I on the right track with the a posteriori analysis?
inferred on the original MSA if I was doing this analysis.
> Should --bootstop-perms and autoMRE be outputting a suggested number
> of suitable BS reps or am I misunderstanding the intention of this?
--bootstop-perms setting is an internal setting for the actual
convergence test that is being performed, and does not set the number of
BS replicates autoMRE will execute before giving up.
Alexis
>
> Thanks!
>
> James
>
> --
> 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>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/raxml/1edc545a-470c-4cb5-b2ec-851ce5c6ce71n%40googlegroups.com
> <https://groups.google.com/d/msgid/raxml/1edc545a-470c-4cb5-b2ec-851ce5c6ce71n%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
Affiliated Scientist, Evolutionary Genetics and Paleogenomics (EGP) lab,
Institute of Molecular Biology and Biotechnology, Foundation for
Research and Technology Hellas
www.exelixis-lab.org
James Munro
Aug 30, 2021, 1:43:30 PM8/30/21
to raxml
Thank you for the insight and advice, Alexis!
James
Reply all
Reply to author
Forward
0 new messages