raxml-ng error: TreeInfo.cpp:270: double TreeInfo::optimize_params(int, double): Assertion `cur_loglh - new_loglh < -new_loglh * RAXML_DOUBLE_TOLERANCE' failed.

[sean...@tmu.edu.tw]

Hi,

I have tried several approaches (both raxml and raxml-ng, see below), but I kept getting errors for some of my files:

1. I used "raxmlHPC-PTHREADS -f a -T 31 -m PROTGAMMALG4X -p 12345 -x 12345 -# 100 -s input.phy"

But I got errors like this:

Bootstrap[96]: Time 1.222145 seconds, bootstrap likelihood -2272.779145, best rearrangement setting 13

Bootstrap[97]: Time 1.186016 seconds, bootstrap likelihood -2208.754567, best rearrangement setting 11

Bootstrap[98]: Time 1.219779 seconds, bootstrap likelihood -2147.585033, best rearrangement setting 6

Bootstrap[99]: Time 1.166089 seconds, bootstrap likelihood -2247.083733, best rearrangement setting 14

raxmlHPC-PTHREADS: optimizeModel.c:1571: optParamGeneric: Assertion `tr->likelihood >= initialLH' failed.

Overall Time for 100 Rapid Bootstraps 127.125646 seconds

Average Time per Rapid Bootstrap 1.271256 seconds

Starting ML Search ...

-2412.8185880121727677760645747184753417968750 -2402.8852359156471720780245959758758544921875

runBacocaResult_reduced_abort.sh: line 26: 23401 Aborted                 (core dumped) raxmlHPC-PTHREADS -f a -T 31 -m PROTGAMMALG4X -p 12345 -x 12345 -# 100 -s input.phy

2. I did some search and found  --no-bfgs might help. However, I got the same error:

raxmlHPC-PTHREADS -f a -T 31  --no-bfgs -m PROTGAMMALG4X -p 12345 -x 12345 -# 100 -s  input.phy

raxmlHPC-PTHREADS: optimizeModel.c:1571: optParamGeneric: Assertion `tr->likelihood >= initialLH' failed.

Overall Time for 100 Rapid Bootstraps 127.125646 seconds

Average Time per Rapid Bootstrap 1.271256 seconds

Starting ML Search ...

-2412.8185880121727677760645747184753417968750 -2402.8852359156471720780245959758758544921875

runBacocaResult_reduced_abort.sh: line 26: 23401 Aborted                 (core dumped) raxmlHPC-PTHREADS -f a -T 31 --no-bfgs -m PROTGAMMALG4X -p 12345 -x 12345 -# 100 -s  input.phy

3. I found raxml-ng might help, but..

a. raxml-ng --all --msa  input.phy --model LG+G4+F --tree pars{20} --bs-trees 200 --threads 1

b. raxml-ng --all --msa  input.phy --model LG+GA --tree pars{20} --bs-trees 200 --threads 1

c. raxml-ng --all --msa  input.phy --model LG+G4 --tree pars{20} --bs-trees 200 --threads 1

a, b, and c have the same error:

Partition 0: noname

Model: LG+FO+G4m

Alignment sites / patterns: 473 / 218

Gaps: 1.10 %

Invariant sites: 10.36 %

[00:00:00] NOTE: Resuming execution from checkpoint (logLH: -4266.21, ML trees: 1, bootstraps: 0)

[00:00:00] Data distribution: partitions/thread: 1-1, patterns/thread: 218-218

Starting ML tree search with 20 distinct starting trees

raxml-ng: /home/alex/projects/hits/raxml-ng/src/TreeInfo.cpp:270: double TreeInfo::optimize_params(int, double): Assertion `cur_loglh - new_loglh < -new_loglh * RAXML_DOUBLE_TOLERANCE' failed.

Aborted (core dumped)

Does anyone know how to solve this problem? Thanks!!

Cheers,

Sean

[Alexey Kozlov]

Hi Sean,

thanks for reporting! I just noticed that with raxml-ng you used a different model: LG+FO+G4 instead of LG4X. Any
specific reason for this?

I can reproduce the error with LG+FO+G4 but not with LG4X or LG+G4.

Best,
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.

[sean...@tmu.edu.tw]

Hi Alexey,

Thanks for your quick response. I tried several models (including LG4X and LG+G4), but all with the same error (I also tried your sever, it generated errors as well), please see the following messages:

1. raxml-ng --all --msa  input.phy --model LG+G4 --tree pars{20} --bs-trees 200 --threads 1

RAxML-NG v. 0.6.0 BETA released on 16.06.2018 by The Exelixis Lab.

Authors: Alexey Kozlov, Alexandros Stamatakis, Diego Darriba, Tomas Flouri, Benoit Morel.

RAxML-NG was called as follows:

raxml-ng --all --msa  input.phy --model LG+G4 --tree pars{20} --bs-trees 200 --threads 1

Analysis options:

  run mode: ML tree search + bootstrapping

  start tree(s): parsimony (20)

  bootstrap replicates: 200

  random seed: 1532916027

  tip-inner: ON

  pattern compression: ON

  per-rate scalers: OFF

  site repeats: OFF

  branch lengths: linked (ML estimate, algorithm: NR-FAST)

  SIMD kernels: AVX

  parallelization: NONE/sequential

WARNING: The model you specified on the command line (LG+G4) will be ignored

         since the binary MSA file already contains a model definition.

         If you want to change the model, please re-run RAxML-NG

         with the original PHYLIP/FASTA alignment and --redo option.

[00:00:00] Loading binary alignment from file: ../OG2419.phy.reduced.raxml.rba

[00:00:00] Alignment comprises 20 taxa, 1 partitions and 218 patterns

Partition 0: noname

Model: LG+FO+G4m

Alignment sites / patterns: 473 / 218

Gaps: 1.10 %

Invariant sites: 10.36 %

[00:00:00] NOTE: Resuming execution from checkpoint (logLH: -4266.21, ML trees: 1, bootstraps: 0)

[00:00:00] Data distribution: partitions/thread: 1-1, patterns/thread: 218-218

Starting ML tree search with 20 distinct starting trees

raxml-ng: /home/alex/projects/hits/raxml-ng/src/TreeInfo.cpp:270: double TreeInfo::optimize_params(int, double): Assertion `cur_loglh - new_loglh < -new_loglh * RAXML_DOUBLE_TOLERANCE' failed.

Aborted (core dumped)

2. raxml-ng --all --msa  input.phy --model LG4X --tree pars{20} --bs-trees 200 --threads 1

RAxML-NG was called as follows:

raxml-ng --all --msa  input.phy --model LG4X --tree pars{20} --bs-trees 200 --threads 1

Analysis options:

  run mode: ML tree search + bootstrapping

  start tree(s): parsimony (20)

  bootstrap replicates: 200

  random seed: 1532916082

  tip-inner: ON

  pattern compression: ON

  per-rate scalers: OFF

  site repeats: OFF

  branch lengths: linked (ML estimate, algorithm: NR-FAST)

  SIMD kernels: AVX

  parallelization: NONE/sequential

WARNING: The model you specified on the command line (LG4X) will be ignored

         since the binary MSA file already contains a model definition.

         If you want to change the model, please re-run RAxML-NG

         with the original PHYLIP/FASTA alignment and --redo option.

[00:00:00] Loading binary alignment from file: ../OG2419.phy.reduced.raxml.rba

[00:00:00] Alignment comprises 20 taxa, 1 partitions and 218 patterns

Partition 0: noname

Model: LG+FO+G4m

Alignment sites / patterns: 473 / 218

Gaps: 1.10 %

Invariant sites: 10.36 %

[00:00:00] NOTE: Resuming execution from checkpoint (logLH: -4266.21, ML trees: 1, bootstraps: 0)

[00:00:00] Data distribution: partitions/thread: 1-1, patterns/thread: 218-218

Starting ML tree search with 20 distinct starting trees

raxml-ng: /home/alex/projects/hits/raxml-ng/src/TreeInfo.cpp:270: double TreeInfo::optimize_params(int, double): Assertion `cur_loglh - new_loglh < -new_loglh * RAXML_DOUBLE_TOLERANCE' failed.

Aborted (core dumped)

3. In both 1 and 2, it said "The model you specified on the command line (LG4X) will be ignored since the binary MSA file already contains a model definition." What does this mean?

4. I also tried your server with LG4X model but it failed. My result is here : https://raxml-ng.vital-it.ch/#/result/675/code/mpQXUkA54G3Y

your job has FAILED

raxml-ng --msa sequenceAlignment.fa --model LG4X --opt-branches on --opt-model on --tree pars{10},rand{10} --all --bs-trees autoMRE{200} --bs-cutoff 0.03 --force --threads 2 --prefix result







Cheers,

Sean

Alexey Kozlov於 2018年7月30日星期一 UTC+8上午7時44分03秒寫道：

[Alexey Kozlov]

Hi Sean,

> 3. In both 1 and 2, it said "The model you specified on the command line (LG4X) will be ignored since the binary MSA
> file already contains a model definition." What does this mean?

This means that tree search was resumed from a checkpoint using the model specified in the very first original run
(LG+FO+G) and not the new model (LG4X etc.). That's why you always got the same error regardless of the specified model.
So please add the "--redo" option as suggested in the warning message.

> 4. I also tried your server with LG4X model but it failed. My result is here :
> https://raxml-ng.vital-it.ch/#/result/675/code/mpQXUkA54G3Y>

For some reason I cannot reproduce this failure locally with the same raxml-ng version. I contacted web server
maintainers and we will figure out what happened there.

Best,
Alexey

[sean...@tmu.edu.tw]

Hi Alexey,

Thank you for your help. However, some of my files now have different errors with LG4X model (below). Please advise. Thanks!

RAxML-NG v. 0.6.0 BETA released on 16.06.2018 by The Exelixis Lab.

RAxML-NG was called as follows:

raxml-ng --all --msa ./input-OG4221.phy --model LG4X --tree pars{20} --bs-trees 200 --threads 1

Analysis options:

  run mode: ML tree search + bootstrapping

  start tree(s): parsimony (20)

  bootstrap replicates: 200

  random seed: 1533229607

  tip-inner: ON

  pattern compression: ON

  per-rate scalers: OFF

  site repeats: OFF

  branch lengths: linked (ML estimate, algorithm: NR-FAST)

  SIMD kernels: AVX

  parallelization: NONE/sequential

[00:00:00] Reading alignment from file: ./input-OG4221.phy

[00:00:00] Loaded alignment with 16 taxa and 376 sites

Alignment comprises 1 partitions and 77 patterns

Partition 0: noname

Model: LG4X+R4

Alignment sites / patterns: 376 / 77

Gaps: 1.08 %

Invariant sites: 13.83 %

NOTE: Binary MSA file created: ./input-OG4221.phy.raxml.rba

[00:00:00] Generating 20 parsimony starting tree(s) with 16 taxa

[00:00:00] Data distribution: partitions/thread: 1-1, patterns/thread: 77-77

Starting ML tree search with 20 distinct starting trees

ERROR: ERROR in branch length optimization (LIBPLL-2240): BL opt converged to a worse likelihood score by -0.000000465680841 units

I am trying to get 31 gene trees and then use these gene trees to obtain a species tree.

However,  26 of the 31 input files have no errors with the LG+GX+F model while  20 of the 31 input files have no errors with the LG4X model. 

I am confused by model selection.... Can I use different models for different genes if I can't get no errors for one model? 

Best,

Sean

[Alexey Kozlov]

Hi Sean,

you should better perform formal modeltesting (e.g. with ModelTest-NG, https://github.com/ddarriba/modeltest) to
determine the optimal model for each gene alignment.

I will have a look at the error.

Best,
Alexey

[sean...@tmu.edu.tw]

Hi Alexey,

Thanks! I tried ModelTest-NG, but there was an error with the latest version. I am still waiting for author's reply.

Btw, have you figured out what happened with the following error in raxml-ng of my previous post? 

 ERROR: ERROR in branch length optimization (LIBPLL-2240): BL opt converged to a worse likelihood score by -0.000000465680841 units

Thanks!

Sean

Alexey Kozlov於 2018年8月3日星期五 UTC+8下午5時05分53秒寫道：

[Alexey Kozlov]

HI Sean,

this seems to be a numerical issue in the branch length optimization, could you please try the following switch:

--blopt nr_safe

This will make tree inference up to 30% slower, but should be more reliable.

Best,

Alexey

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