Bootstopping - continue bootstrapping beyond 50 bootstraps
48 views
Skip to first unread message
Ales Bucek
Jul 27, 2017, 11:05:11 PM7/27/17
to raxml
I'd like to ask for a help with hopefully trivial question concerning bootstopping.
I run RAxML 8.2.4 as follows
But either using autoMR, as above, or using autoMRE, the run stops after 50 bootstraps. As I understood the functionality of bootstopping it should continue until a WFR threshold is reached. See attached the info file outputted by RAxML.
Thank you in advance for any hints.
Best,
Ales
I run RAxML 8.2.4 as follows
raxmlHPC-PTHREADS-SSE3 -T 24 -f a -s FcC_smatrix.phy -x 6907 -# autoMR -m GTRGAMMA -n raxml.out -q ps.txt -p 4893Thank you in advance for any hints.
Best,
Ales
Alexandros Stamatakis
Jul 28, 2017, 3:12:19 AM7/28/17
to ra...@googlegroups.com
Hi Alexs,
That is really weird, could you check if this also happens with the
latest release version 8.2.11
https://github.com/stamatak/standard-RAxML/releases
if the problem persists then please send me your input files to my
personal email.
All the best,
Alexis
On 28.07.2017 05:05, Ales Bucek wrote:
> I'd like to ask for a help with hopefully trivial question concerning
> bootstopping.
>
> I run RAxML 8.2.4 as follows
> |
> raxmlHPC-PTHREADS-SSE3 -T 24-f a -s FcC_smatrix.phy -x 6907-# autoMR -m
> 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
www.exelixis-lab.org
That is really weird, could you check if this also happens with the
latest release version 8.2.11
https://github.com/stamatak/standard-RAxML/releases
if the problem persists then please send me your input files to my
personal email.
All the best,
Alexis
On 28.07.2017 05:05, Ales Bucek wrote:
> I'd like to ask for a help with hopefully trivial question concerning
> bootstopping.
>
> I run RAxML 8.2.4 as follows
> |
> GTRGAMMA -n raxml.out -q ps.txt -p 4893
> |
> But either using autoMR, as above, or using autoMRE, the run stops after
> 50 bootstraps. As I understood the functionality of bootstopping it
> should continue until a WFR threshold is reached. See attached the info
> file outputted by RAxML.
> Thank you in advance for any hints.
>
> Best,
> Ales
>
> --
> |
> But either using autoMR, as above, or using autoMRE, the run stops after
> 50 bootstraps. As I understood the functionality of bootstopping it
> should continue until a WFR threshold is reached. See attached the info
> file outputted by RAxML.
> Thank you in advance for any hints.
>
> Best,
> Ales
>
> 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
www.exelixis-lab.org
Ales Bucek
Jul 31, 2017, 9:56:28 AM7/31/17
to raxml
Hi Alexis,
thanks for your quick response. I rerun the analysis using v. 8.2.11 and command
with the same result as previously (rapid BS search stopped after 50 BS replicates). I'm sending to your email the input data.
Thanks for your help!
Best,
Ales
thanks for your quick response. I rerun the analysis using v. 8.2.11 and command
raxmlHPC-PTHREADS-AVX2 -T 24 -f a -s FcC_smatrix.phy -x 6907 -# autoMRE -m GTRGAMMA -n raxml.out -q codonpartition.txt -p 4893Thanks for your help!
Best,
Ales
Alexandros Stamatakis
Aug 1, 2017, 3:06:48 AM8/1/17
to ra...@googlegroups.com
Dear Ales,
The problem was that RAxML did not report in which units this is measured.
It actually reports the average relative WRF in per cent!
Thus, the printout:
Stopped Rapid BS search after 50 replicates with MR-based Bootstopping
criterion
WRF Average of 100 random splits: 0.520615
means that the WRF is on average 0.520615 % i.e., very small.
I've just pushed an improved printout where the % is actually printed to
make this more clear.
Alexis
On 28.07.2017 05:05, Ales Bucek wrote:
The problem was that RAxML did not report in which units this is measured.
It actually reports the average relative WRF in per cent!
Thus, the printout:
Stopped Rapid BS search after 50 replicates with MR-based Bootstopping
criterion
WRF Average of 100 random splits: 0.520615
means that the WRF is on average 0.520615 % i.e., very small.
I've just pushed an improved printout where the % is actually printed to
make this more clear.
Alexis
On 28.07.2017 05:05, Ales Bucek wrote:
> I'd like to ask for a help with hopefully trivial question concerning
> bootstopping.
>
> I run RAxML 8.2.4 as follows
> |
> raxmlHPC-PTHREADS-SSE3 -T 24-f a -s FcC_smatrix.phy -x 6907-# autoMR -m
> bootstopping.
>
> I run RAxML 8.2.4 as follows
> |
> GTRGAMMA -n raxml.out -q ps.txt -p 4893
> |
> But either using autoMR, as above, or using autoMRE, the run stops after
> 50 bootstraps. As I understood the functionality of bootstopping it
> should continue until a WFR threshold is reached. See attached the info
> file outputted by RAxML.
> Thank you in advance for any hints.
>
> Best,
> Ales
>
> |
> But either using autoMR, as above, or using autoMRE, the run stops after
> 50 bootstraps. As I understood the functionality of bootstopping it
> should continue until a WFR threshold is reached. See attached the info
> file outputted by RAxML.
> Thank you in advance for any hints.
>
> Best,
> Ales
>
Ales Bucek
Aug 4, 2017, 2:43:42 AM8/4/17
to raxml
Dear Alexis,
many thanks for the clarification!
Interpreting the reported WRF as a percentage absolutely did not come to my mind as I was not expecting that 50 BS replicates might reach the default WRF threshold for all the datasets for which I run RAxML (although these datasets are actually one really large and then various smaller subsets of the large dataset so they might be expected to perform in a similar way) - still I consider it a bit unexpected result but apparently this might be the same case as with your experimental dataset "125" from your 2010 paper on Bootstopping criteria.
Best,
Ales
many thanks for the clarification!
Interpreting the reported WRF as a percentage absolutely did not come to my mind as I was not expecting that 50 BS replicates might reach the default WRF threshold for all the datasets for which I run RAxML (although these datasets are actually one really large and then various smaller subsets of the large dataset so they might be expected to perform in a similar way) - still I consider it a bit unexpected result but apparently this might be the same case as with your experimental dataset "125" from your 2010 paper on Bootstopping criteria.
Best,
Ales
Alexandros Stamatakis
Aug 4, 2017, 2:46:56 AM8/4/17
to ra...@googlegroups.com
Dear Ales,
We have repeatedly observed that for broad phylogenomic datasets
(relatively few taxa with hundreds or thousands of genes or even whole
genomes/transcriptomes) the bootstrap converges after 50 replicates.
Evidently this does not need to be always the case, but there is a clear
trend.
If you want to double check you can use RAxML to compute all pair-wise
RF distances between those 50 bootstrap trees to see how different they
are topologically.
Alexis
> > an email to raxml+un...@googlegroups.com <javascript:>
> > <mailto:raxml+un...@googlegroups.com <javascript:>>.
We have repeatedly observed that for broad phylogenomic datasets
(relatively few taxa with hundreds or thousands of genes or even whole
genomes/transcriptomes) the bootstrap converges after 50 replicates.
Evidently this does not need to be always the case, but there is a clear
trend.
If you want to double check you can use RAxML to compute all pair-wise
RF distances between those 50 bootstrap trees to see how different they
are topologically.
Alexis
> > <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
>
> www.exelixis-lab.org <http://www.exelixis-lab.org>
> --
> Alexandros (Alexis) Stamatakis
>
> Research Group Leader, Heidelberg Institute for Theoretical Studies
> Full Professor, Dept. of Informatics, Karlsruhe Institute of Technology
>
Reply all
Reply to author
Forward
0 new messages