Can I Map Bootstrap to Separately Generated ML Tree?
xiaoyu zhao
Dear RAxML development team,
I’m currently using RAxML for phylogenetic analysis and encountered a situation I would appreciate your guidance on.
Initially, I ran the following command to perform both bootstrap analysis and ML tree search :
The program successfully completed the 100 bootstrap replicates and generated the RAxML_bootstrap file. However, during the ML tree search stage, the process was terminated (killed), so I was only able to obtain the bootstrap trees but not the final ML tree (RAxML_bestTree).
To avoid rerunning the full analysis, I plan to:
-
Perform a separate ML tree search using：raxmlHPC-PTHREADS \
-T 64 \ -p 12345 \ -m GTRCAT \ -s merge_sort_biallelic_filtered.recode.min4.phy \ -n mihou_filter_ml Map the bootstrap support values using -f b with the generated ML tree and the existing RAxML_bootstrap file. raxmlHPC-PTHREADS \ -f b \ -t RAxML_bestTree.mihou_filter_ml \ -z RAxML_bootstrap.mihou_filter \
-m GTRCAT \
-n mihou_filter_final
My questions are:
-
Is it scientifically valid to separately compute the ML tree after bootstrap and then use -f b to annotate it with bootstrap support?
-
Would this approach be equivalent to the -f a result in terms of reliability?
I greatly appreciate your time and any advice you can provide.
Alexandros Stamatakis
This will be perfectly fine, both scientifically and in terms of
expected quality of results.
However, I would execute more than one ML tree search, i.e., 20
independent tree searches.
Also, I would strongly recommend that you use the up to date version of
RAxML: RAxML-NG https://github.com/amkozlov/raxml-ng
Your thread setting seems to be a bit high, RAxML-NG will figure this
out automatically for you. Also, the fact that the program didn't
terminate may indicate that it ran out of memory during the ML search,
so I assume that you must be analyzing a very large dataset.
Alexis
On 24.04.25 04:29, xiaoyu zhao wrote:
> Dear RAxML development team,
>
> I’m currently using RAxML for phylogenetic analysis and encountered a
> situation I would appreciate your guidance on.
>
> Initially, I ran the following command to perform both bootstrap
> analysis and ML tree search :
>
> The program successfully completed the 100 bootstrap replicates and
> generated the RAxML_bootstrap file. However, during the ML tree search
> stage, the process was terminated (killed), so I was only able to obtain
> the bootstrap trees but not the final ML tree (RAxML_bestTree).
>
> To avoid rerunning the full analysis, I plan to:
>
>
> Perform a separate ML tree search using：*raxmlHPC-PTHREADS \*
>
> * -T 64 \ -p 12345 \ -m GTRCAT \ -s
> merge_sort_biallelic_filtered.recode.min4.phy \ -n mihou_filter_ml*
> 2.
> Map the bootstrap support values using -f b with the generated ML
> b \ -t RAxML_bestTree.mihou_filter_ml \ -z
> RAxML_bootstrap.mihou_filter \*
>
> * -m GTRCAT \
> -n mihou_filter_final*
>
> My questions are:
>
> *
> Is it scientifically valid to separately compute the ML tree after
> bootstrap and then use -f b to annotate it with bootstrap support?
>
> Would this approach be equivalent to the -f a result in terms of
> reliability?
>
> I greatly appreciate your time and any advice you can provide.
>
> 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 visit
> https://groups.google.com/d/msgid/raxml/89a88c10-208e-4cd6-88e4-9235d6867099n%40googlegroups.com <https://groups.google.com/d/msgid/raxml/89a88c10-208e-4cd6-88e4-9235d6867099n%40googlegroups.com?utm_medium=email&utm_source=footer>.
--
Alexandros (Alexis) Stamatakis
ERA Chair, Institute of Computer Science, Foundation for Research and
Technology - Hellas
Research Group Leader, Heidelberg Institute for Theoretical Studies
Full Professor, Dept. of Informatics, Karlsruhe Institute of Technology
www.biocomp.gr (Crete lab)
www.exelixis-lab.org (Heidelberg lab)