Error when starting ML search: "raxmlHPC-HYBRID: optimizeModel.c:1571: optParamGeneric: Assertion `tr->likelihood >= initialLH' failed."

[Andrew Young]

Hello, 

I'm currently trying to run RAxML on CIPRES with a hybrid enrichment amino acid dataset (partitions determined in PF2).

After bootstrapping, I'm getting the following error when RAxML attempts to start the ML search:

raxmlHPC-HYBRID: optimizeModel.c:1571: optParamGeneric: Assertion `tr->likelihood >= initialLH' failed.
[comet-26-16:18744] *** Process received signal ***
[comet-26-16:18744] Signal: Aborted (6)
[comet-26-16:18744] Signal code:  (-6)
[comet-26-16:18744] [ 0] /lib64/libpthread.so.0[0x3eae00f7e0]
[comet-26-16:18744] [ 1] /lib64/libc.so.6(gsignal+0x35)[0x3ead832495]
[comet-26-16:18744] [ 2] /lib64/libc.so.6(abort+0x175)[0x3ead833c75]
[comet-26-16:18744] [ 3] /lib64/libc.so.6[0x3ead82b60e]
[comet-26-16:18744] [ 4] /lib64/libc.so.6(__assert_perror_fail+0x0)[0x3ead82b6d0]
[comet-26-16:18744] [ 5] raxmlHPC-HYBRID[0x439d45]
[comet-26-16:18744] [ 6] raxmlHPC-HYBRID[0x442ebe]
[comet-26-16:18744] [ 7] raxmlHPC-HYBRID[0x437080]
[comet-26-16:18744] [ 8] raxmlHPC-HYBRID[0x448431]
[comet-26-16:18744] [ 9] raxmlHPC-HYBRID[0x404c4d]
[comet-26-16:18744] [10] /lib64/libc.so.6(__libc_start_main+0xfd)[0x3ead81ed1d]
[comet-26-16:18744] [11] raxmlHPC-HYBRID[0x402d59]
[comet-26-16:18744] *** End of error message ***

CIPRES is calling RAxML using the following command line:

raxmlHPC-HYBRID -T 4 -n result --no-bfgs -s infile.txt -q part.txt -p 12345 -m PROTGAMMAWAG -f a -N 10 -x 12345 

And these are the last lines in the STOUT file where the program crashes:

Starting ML Search ...

-1861334.2922942275181412696838378906250000000000 -1860796.0279493641573935747146606445312500000000
-1861339.0463542430661618709564208984375000000000 -1860796.4388570610899478197097778320312500000000
-1861320.8045942354947328567504882812500000000000 -1860781.5823094076476991176605224609375000000000
-1861334.2663305485621094703674316406250000000000 -1860795.9613526051398366689682006835937500000000
-1861320.7807515002787113189697265625000000000000 -1860781.5025298362597823143005371093750000000000

Does anybody know what's causing this? Any help would be very much appreciated.

Thank you, 

Andrew

[Alexandros Stamatakis]

This looks like a numerical issue, my advice would be that you try using
RAxML-NG instead: https://github.com/amkozlov/raxml-ng

Alexis

> --
> 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

[Andrew Young]

Hello Dr. Stamatakis, 

Thank you very much for the advice!  However, I just looked through RAxML-ng's help output, and I don't see anything about partitioned analyses.  Is this possible with the current build of NG and I'm just missing something?

Thanks again, 

Andrew

[Alexey Kozlov]

Hi Andrew,

yes it is possible, see details here (please note the subtle differences to RAxML partition file!):

https://github.com/amkozlov/raxml-ng/wiki/Input-data#multiple-models

Best,
Alexey

On 14.12.2017 16:50, Andrew Young wrote:
> Hello Dr. Stamatakis,
>
> Thank you very much for the advice!  However, I just looked through RAxML-ng's help output, and I don't see anything
> about partitioned analyses.  Is this possible with the current build of NG and I'm just missing something?
>
> Thanks again,
>
> Andrew
>
>
>
> On Thursday, December 14, 2017 at 4:21:14 AM UTC-5, Alexis wrote:
>
> This looks like a numerical issue, my advice would be that you try using

> RAxML-NG instead: https://github.com/amkozlov/raxml-ng <https://github.com/amkozlov/raxml-ng>

> > an email to raxml+un...@googlegroups.com <javascript:>
> > <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>

[Andrew Young]

Hi again, 

sorry I missed that option in the documentation.  It seems to be working, but I don't have access to a cluster at the moment (aside from CIPRES) so I'm running it locally on a spare iMac.  Hopefully it can handle it if I leave it running for several days/weeks.

Thanks again for the help.

Cheers, 

Andrew

[Alexey Kozlov]

You're welcome, good luck with your analysis! :)

> > www.exelixis-lab.org <http://www.exelixis-lab.org> <http://www.exelixis-lab.org>

> >
> > --
> > 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

> <javascript:>
> > <mailto:raxml+un...@googlegroups.com <javascript:>>.
> > For more options, visit https://groups.google.com/d/optout <https://groups.google.com/d/optout>.
>
> --