Raxml-ng terminated with error

[fe...@siswa.um.edu.my]

Hi,

I am running a concatenated gene alignment (78 genes) of 114 bacterial genome using RAxML-NG v. 0.7.0git BETA. I am using ModelTest-ng for partitioning. The command that I used was as below:

./raxml-ng --all --msa /home/umhir/Desktop/concatenated_core_gene_alignments_Planococcaceae_trimal.fasta --msa-format FASTA --data-type DNA --seed 100000 --threads 20 --brlen unlinked --model /home/umhir/Desktop/Partition_Planococcaceae_modeltest.txt --bs-trees 100

The run terminated after a long running time, and I got error as below:

terminate called after throwing an instance of 'std::runtime_error'

   what(): ERROR in SPR round (LIBPLL-2240): BL opt converged to a worse likelihood score by -0.000020038336515 units

Aborted (core dumped)

Please find attached the log file. 

Please advice what should I do?

Best regards,

Felix

[Alexey Kozlov]

Hi Felix,

thanks for reporting, I have few suggestions:

1) please try the latest github version, I recently fixed a bug which
caused similar symptoms

2) use "--blopt nr_safe" as a quick workaround

3) please send me your input files (alignment+partitions) such the I can
investigate the problem

Hope this helps,
Alexey

> --
> 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.

[fe...@siswa.um.edu.my]

Dear Alexey,

Thanks for you reply. I thought that v0.7.0 is the latest version. I will try it as soon as I can.

Please find attached my alignment and partition file. Thank you.

Best regards,

Felix

[fe...@siswa.um.edu.my]

Hi Alexey,

May I know which version is the latest version? I only found version v0.7.0git BETA that I am using now. Thank you.

Best regards,

Felix

[SEE TOO WAH SENG STUDENT]

Hi Alexey,

After added --blopt nr_safe command, I got error at the "Starting ML tree search" step, and the error is as below:

raxml-ng: /home/umhir/raxml-ng/libs/pll-modules/src/optimize/pll_optimize.c:1880: pllmod_opt_optimize_branch_lengths_local_multi: Assertion `new_loglikelihood - loglikelihood > new_loglikelihood * BETTER_LL_TRESHOLD' failed.

raxml-ng: /home/umhir/raxml-ng/libs/pll-modules/src/optimize/pll_optimize.c:1880: pllmod_opt_optimize_branch_lengths_local_multi: Assertion `new_loglikelihood - loglikelihood > new_loglikelihood * BETTER_LL_TRESHOLD' failed.

raxml-ng: /home/umhir/raxml-ng/libs/pll-modules/src/optimize/pll_optimize.c:1880: pllmod_opt_optimize_branch_lengths_local_multi: Assertion `new_loglikelihood - loglikelihood > new_loglikelihood * BETTER_LL_TRESHOLD' failed.

raxml-ng: /home/umhir/raxml-ng/libs/pll-modules/src/optimize/pll_optimize.c:1880: pllmod_opt_optimize_branch_lengths_local_multi: Assertion `new_loglikelihood - loglikelihood > new_loglikelihood * BETTER_LL_TRESHOLD' failed.

Aborted (core dumped)

Please advice.

Best regards,

Felix

To unsubscribe from this group and stop receiving emails from it, send an email to raxml+un...@googlegroups.com.

[Alexey Kozlov]

Hi Felix,

you need the version from 31.10.2018, i.e. you should see the following
line in the output:

RAxML-NG v. 0.7.0git BETA released on 31.10.2018 by The Exelixis Lab.

Best,
Alexey

On 03.11.18 09:26, fe...@siswa.um.edu.my wrote:
> Hi Alexey,
>
> May I know which version is the latest version? I only found version
> v0.7.0git BETA that I am using now. Thank you.
>
>
>
> Best regards,
>
> Felix
>
> On Friday, November 2, 2018 at 10:03:09 PM UTC+8, Alexey Kozlov wrote:
>
> Hi Felix,
>
> thanks for reporting, I have few suggestions:
>
> 1) please try the latest github version, I recently fixed a bug which
> caused similar symptoms
>
> 2) use "--blopt nr_safe" as a quick workaround
>
> 3) please send me your input files (alignment+partitions) such the I
> can
> investigate the problem
>
> Hope this helps,
> Alexey
>

> > 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>.

[SEE TOO WAH SENG STUDENT]

Hi Alexey,

Thanks. Then I am actually using the latest version.

Best regards,

Felix

To unsubscribe from this group and stop receiving emails from it, send an email to raxml+un...@googlegroups.com.

[KW]

Hi, 

I am encountering a similar problem in which I get an error message: RAXML_DOUBLE_TOLERANCE...

I am estimating a tree for a very small SNP dataset and using an ascertainment bias correction. I have tried the  --blopt nr_safe argument and am attaching the multiple alignment. I am using this version: 

RAxML-NG v. 0.8.1 BETA released on 05.03.2019 by The Exelixis Lab.

```

raxml-ng: /home/alex/hits/raxng-release/src/TreeInfo.cpp:275: double TreeInfo::optimize_params(int, double): Assertion `cur_loglh - new_loglh < -new_loglh * RAXML_DOUBLE_TOLERANCE' failed.

/var/spool/slurm/slurmd/job8811626/slurm_script: line 59: 153162 Aborted                 (core dumped) raxml-ng --all --msa ${fasta} --model ${model} --prefix ${fasta/_snps.fa}_${model} --threads 1 --blopt nr_safe +ASC_STAM{$A_invar,$C_invar,$G_invar,$T_invar}

```

My second question is about the ascertainment bias correction. If I use the ASC_STAM correction and specify the number of invariant sites, is it also necessary to specify the proportion of invariant sites with the +I argument?

Thank you in advance!

Best,

> an email to ra...@googlegroups.com
> <mailto:ra...@googlegroups.com>.

> For more options, visit https://groups.google.com/d/optout.

--
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 ra...@googlegroups.com.

[Alexey Kozlov]

Hi Katharine,

> I am encountering a similar problem in which I get an error message:
> RAXML_DOUBLE_TOLERANCE...
> I am estimating a tree for a very small SNP dataset and using an
> ascertainment bias correction. I have tried the --blopt nr_safe argument
> and am attaching the multiple alignment. I am using this version:

could you please send me your input files and full raxml-ng output?

> My second question is about the ascertainment bias correction. If I use
> the ASC_STAM correction and specify the number of invariant sites, is it
> also necessary to specify the proportion of invariant sites with the +I
> argument?

No, since you don't have any invariable sites in your MSA, you should
not use +I model

Best,
Alexey

>
> Thank you in advance!
> Best,
>
> On Saturday, November 3, 2018 at 6:26:53 AM UTC-7, SEE TOO WAH SENG
> STUDENT wrote:
>
> Hi Alexey,
>
> Thanks. Then I am actually using the latest version.
>
> Best regards,
>
> Felix
>
>
>
> On Sat, Nov 3, 2018 at 8:55 PM Alexey Kozlov <alexei...@gmail.com
> <javascript:>> wrote:
>
> Hi Felix,
>
> you need the version from 31.10.2018, i.e. you should see the
> following
> line in the output:
>
> RAxML-NG v. 0.7.0git BETA released on 31.10.2018 by The Exelixis
> Lab.
>
> Best,
> Alexey
>

> > an email to ra...@googlegroups.com <javascript:>
> > <mailto:ra...@googlegroups.com <javascript:>>.

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

> For more options, visit https://groups.google.com/d/optout
> <https://groups.google.com/d/optout>.
>
> --
> 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>.

[KW]

Thank you Alexey for the quick reply. 

I am attaching the MSA and RAxML log file. 

The error below is output to the screen but not the RAxML log file. 

raxml-ng: /home/alex/hits/raxng-release/src/TreeInfo.cpp:275: double TreeInfo::optimize_params(int, double): Assertion `cur_loglh - new_loglh < -new_loglh * RAXML_DOUBLE_TOLERANCE' failed.

Aborted (core dumped)

I am wondering if the issue is related to trying to fit a tree to an alignment with so much missing data. Are there best practices for fitting ML trees with a small number of taxa (17) and a small number of SNP variants (~25)? 

Thank you again. 

Best,

> an email to ra...@googlegroups.com
> <mailto:ra...@googlegroups.com>.

[Alexey Kozlov]

Thanks for the files!

First of all, you must specify asc. bias correction as part of the model
definition, i.e.

--model HKY+ASC_STAM{758551/1449996/1444613/758365}

With your command line, raxml-ng was using plain HKY model, as you can
see in the log file:

Partition 0: noname
Model: HKY+FO
Alignment sites / patterns: 13 / 13
Gaps: 49.77 %
Invariant sites: 0.00 %

In general, this kind of datasets is very problematic for ML tree
inference, since you have so few sites and on top of that 50% missing
data. Unfortunately, there is not much you can do in terms of raxml-ng
settings to improve the situation.

There are a couple of options you could try *for exploratory purposes*,
since both can introduce huge bias and can *not* be considered best
practice:

- use pre-defined nucleotide frequencies estimated from a larger
alignment: "+F{fA/fC/fG/fT}"

- use a constraint tree obtained from a larger alignment:
"--tree-constraint tree.nw"

Best,
Alexey

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

[Katharine Walter]

Dear Alexey,

Thank you for this. I had used the ASC correction argument incorrectly. I will test out the options you proposed. 

Thanks!

To unsubscribe from this group and stop receiving emails from it, send an email to raxml+un...@googlegroups.com.