There now is an entire zoo with fast support measures, e.g., aLRT test and derivatives thereof, ultrafast bootstrap (probably not applicable due to the nature of the search algorithm), quartet-based measures: We need to decide at some point which ones we want to integrate.

Hi there

I found a small bug when using partition files that contain a blank line at the end. raxml-ng terminates with an uninformative error message. While fixing the partition file is fine, it would be helpful if the error message hinted at a malformed partition file. A blank line in the partition file may not be obvious to the user. Apologies in advance if I missed/omitted something.

RAxML-NG v. 0.5.0 BETA released on 10.09.2017 by The Exelixis Lab.
Authors: Alexey Kozlov, Alexandros Stamatakis, Diego Darriba, Tomas Flouri, Benoit Morel.
Latest version: https://github.com/amkozlov/raxml-ng
Questions/problems/suggestions? Please visit: https://groups.google.com/forum/#!forum/raxml

WARNING: This is a BETA release, please use at your own risk!

RAxML-NG was called as follows:

raxml-ng-mpi --threads 4 --evaluate --msa ./RT.B.DE.nuc.aligned.fixed.fsa --tree RTIQ.bllenfix.nuc.treefile --model rt.b.de.raxmlng.part --brlen scaled --prefix rt.b.de.nuc.raxml

Analysis options:
run mode: Evaluate tree likelihood
start tree(s): user
random seed: 1513523851
tip-inner: ON
pattern compression: ON
per-rate scalers: OFF
site repeats: OFF
branch lengths: ML estimate (proportional)
SIMD kernels: AVX2
parallelization: PTHREADS (4 threads)

[00:00:00] Reading alignment from file: ./RT.B.DE.nuc.aligned.fixed.fsa
[00:00:00] Loaded alignment with 23403 taxa and 1377 sites

ERROR: h6�

Hi
When I set this dataset as the input alignment, link
I got different loglikelihood from different versions. The model I used:
PTHREADS: --msa alignment.phy --model GTR+G+F
MPI: --msa alignment.phy --model GTR+G+F --threads 1
In PTHREADS version, the loglikehood is about -14406, while in MPI version, the loglikelihood is about -14018. I have tested several times but got similar results. I am just wondering why this is happening.
Thanks in advance.

The function of morphological weight calibration is available in the standard raxml. Is there any adequate option available in raxml-ng?

Hi,

I'm currently trying to build the mpi version of raxml.

I got the following error:

/usr/bin/ld: /home/dhb13/source/raxml-ng-mpi/localdeps/lib/libpllmodalgorithm_static.a(pllmod_algorithm.c.o): unrecognized relocation (0x2a) in section `.text'
/usr/bin/ld: final link failed: Bad value
collect2: error: ld returned 1 exit status
make[2]: *** [../bin/raxml-ng-mpi] Error 1
make[1]: *** [src/CMakeFiles/raxml_module.dir/all] Error 2
make: *** [all] Error 2

I'm using gcc 6.2 and openmpi/3.1.0.

Do you have any suggestions for how to fix this error?

Since raxml-ng relies on libpll, maybe this feature request is best lodged there rather than here. A quick look shows a similar feature request open since 2013 but presumably for the KNC add-in card and the conclusion was "we should not do this on the current code from the master branch." https://app.assembla.com/spaces/phylogenetic-likelihood-library/tickets/72-port-likelihood-function-xeon-phi/details#

So I think here might be the best place?

Rationale for this request is both personal and general -- personally, I have observed ExaML-KNL is up to 5x faster than NG on my KNL machine; generally, we will all soon have AVX512 anyway! :)

(RAxML -M option)

Might require some changes to the llpll, see related paper here: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4604832/

Hello,
I tried to run RAxML-NG using a partitioned dataset with models defined by ModelTest-NG. However, RAxML-NG stopped prematurely (see the attached file). I tried to compile the tool myself and get almost the same error; only the path was different.

Thank you!
screenlog.txt

I believe this is an important option that is rather frequently used.

Hello,
I'm getting the following errors when compile the software:
ParallelContext.hpp:32:61: error: ‘function’ in namespace ‘std’ does not name a template type
static void init_pthreads(const Options& opts, const std::function<void()>& thread_main);
ParallelContext.hpp:32:69: error: expected ‘,’ or ‘...’ before ‘<’ token
static void init_pthreads(const Options& opts, const std::function<void()>& thread_main);
ParallelContext.hpp:45:38: error: ‘std::function’ has not been declared
static void mpi_gather_custom(std::function<int(void*,int)> prepare_send_cb,
ParallelContext.hpp:45:46: error: expected ‘,’ or ‘...’ before ‘<’ token
static void mpi_gather_custom(std::function<int(void*,int)> prepare_send_cb,
ParallelContext.hpp:83:57: error: ‘function’ in namespace ‘std’ does not name a template type
static void start_thread(size_t thread_id, const std::function<void()>& thread_main);
ParallelContext.hpp:83:65: error: expected ‘,’ or ‘...’ before ‘<’ token
static void start_thread(size_t thread_id, const std::function<void()>& thread_main);
Checkpoint.cpp:165:58: error: no matching function for call to ‘ParallelContext::mpi_gather_custom(CheckpointManager::gather_model_params()::<lambda(void*, size_t)>&, CheckpointManager::gather_model_params()::<lambda(void*, size_t)>&)’
ParallelContext::mpi_gather_custom(worker_cb, master_cb);

Can you maybe give me a hint about what I'm doing wrong?
Thank you
Alex

Hello,

Thanks for developing and maintaining this fascinating method.

I want to search an ML tree in my population data. I know it will be very difficult to converge into a bifurcating tree, but I only want to learn about large branches currently, and not hoping to spend a large chunk of time to "optimize" tips. I notice the likelihood value increased quickly and keep increasing just a slight bit (but taking a couple hours each round). In this case, can I stop the run and get the current tree? Or is that any way to set a likelihood cutoff to stop the run? I notice the "--spr-cutoff" parameter but not quite understand what it means and how to set a proper value.

Below is the command I used.

raxml-ng --msa Chr1.imputed.fa --model GTR+G --threads 50 --seed 12315

Any suggestions are welcome. Thanks ahead!

Shujun

I.e., not repeats, but missing data, terraces, trading computation fro memory (Fernando's old paper).

Description: https://www.biorxiv.org/content/early/2017/06/23/154542

Implementation: https://github.com/evolbioinfo/booster

best before: mid-April

I can't seem to compile raxml-ng against any version of libpll. I have tried 0.3.2, 0.3.0 and 0.2.0. which version do you compile against ???

libpll provides only one header file. common.h is calling a bunch.
a number of constants aren't defined anywhere in libpll, such as PLLMOD_OPT_MAX_ALPHA

I have tried 0.3.2, 0.3.0 and 0.2.0. which version do you compile against ???

Wondering if you can help with this error after running make on HPC raxml-ng_v0

$ make
Scanning dependencies of target raxml_module
[  3%] Building CXX object src/CMakeFiles/raxml_module.dir/Checkpoint.cpp.o
[  7%] Building CXX object src/CMakeFiles/raxml_module.dir/CommandLineParser.cpp.o
[ 10%] Building CXX object src/CMakeFiles/raxml_module.dir/LoadBalancer.cpp.o
[ 14%] Building CXX object src/CMakeFiles/raxml_module.dir/MSA.cpp.o
[ 17%] Building CXX object src/CMakeFiles/raxml_module.dir/Model.cpp.o
[ 21%] Building CXX object src/CMakeFiles/raxml_module.dir/Optimizer.cpp.o
[ 25%] Building CXX object src/CMakeFiles/raxml_module.dir/Options.cpp.o
[ 28%] Building CXX object src/CMakeFiles/raxml_module.dir/ParallelContext.cpp.o
[ 32%] Building CXX object src/CMakeFiles/raxml_module.dir/PartitionAssignment.cpp.o
[ 35%] Building CXX object src/CMakeFiles/raxml_module.dir/PartitionInfo.cpp.o
[ 39%] Building CXX object src/CMakeFiles/raxml_module.dir/PartitionedMSA.cpp.o
[ 42%] Building CXX object src/CMakeFiles/raxml_module.dir/SystemTimer.cpp.o
[ 46%] Building CXX object src/CMakeFiles/raxml_module.dir/Tree.cpp.o
[ 50%] Building CXX object src/CMakeFiles/raxml_module.dir/TreeInfo.cpp.o
[ 53%] Building CXX object src/CMakeFiles/raxml_module.dir/log.cpp.o
[ 57%] Building CXX object src/CMakeFiles/raxml_module.dir/main.cpp.o
[ 60%] Building CXX object src/CMakeFiles/raxml_module.dir/sysutil.cpp.o
[ 64%] Building CXX object src/CMakeFiles/raxml_module.dir/io/NewickStream.cpp.o
[ 67%] Building CXX object src/CMakeFiles/raxml_module.dir/io/RBAStream.cpp.o
[ 71%] Building CXX object src/CMakeFiles/raxml_module.dir/io/binary_io.cpp.o
[ 75%] Building CXX object src/CMakeFiles/raxml_module.dir/io/msa_streams.cpp.o
[ 78%] Building CXX object src/CMakeFiles/raxml_module.dir/io/part_info.cpp.o
[ 82%] Building CXX object src/CMakeFiles/raxml_module.dir/bootstrap/BootstopCheck.cpp.o
[ 85%] Building CXX object src/CMakeFiles/raxml_module.dir/bootstrap/BootstrapGenerator.cpp.o
[ 89%] Building CXX object src/CMakeFiles/raxml_module.dir/bootstrap/BootstrapTree.cpp.o
[ 92%] Building CXX object src/CMakeFiles/raxml_module.dir/autotune/ResourceEstimator.cpp.o
[ 96%] Building CXX object src/CMakeFiles/raxml_module.dir/terraces/TerraceWrapper.cpp.o
[100%] Linking CXX executable ../../bin/raxml-ng-mpi
/usr/bin/ld: BFD version 2.20.51.0.2-5.47.el6_9.1 20100205 internal error, aborting at reloc.c line 443 in bfd_get_reloc_size

/usr/bin/ld: Please report this bug.

collect2: error: ld returned 1 exit status
make[2]: *** [../bin/raxml-ng-mpi] Error 1
make[2]: *** Deleting file `../bin/raxml-ng-mpi'
make[1]: *** [src/CMakeFiles/raxml_module.dir/all] Error 2
make: *** [all] Error 2

Similar to "-f c" in RAxML8

A lot of discrete tree operations (IC/TC scores, tree plausibility analysis) and some new stuff (quartets and this paper here http://www.biorxiv.org/content/biorxiv/early/2017/06/23/154542.full.pdf) should not go into raxml-ng but rather to the modules. We need to discuss this at some point, but I think all operations on a discrete set of trees that don't require likelihood calculations should not be part of RAxML-NG.

-g option in RAxML

this is for a specific use-case, i.e., placing a couple of potential outgroups (genome-size sequences) into an ingroup phylogeny, we don't need it super-urgently nor do we need it to be super-efficient, but something like within the next 6 months would be nice

This was a recurrent issue in RAxML when dataset had so little diversity that the br-len optimization hit the branch length minimum set in the code. Also need to be careful to maybe resolve inner nodes with zero branch lengths into multi-furcations. There were 1-2 papers discussing this.

Hello! I am trying to create a tree using raxml-ng on an alignment (99322 taxa and 1285 sites) using

./raxml-ng --msa alignment_pfiltered.fasta --model GTR+G4 --search --threads 45 --tree rand{20}

Unfortunately, I run into the following error:

terminate called after throwing an instance of 'std::runtime_error'
  what():  ERROR in branch lenght optimization: wrong likelihood derivatives

Any guidance on what I'm missing or how I might troubleshoot?

To be discussed I think, for pedagogical reasons as this is only a drawing option maybe we should not add this. It adds quite some code without any added value.

e.g., C10-C60 and other protein mixture models?

PHYLIP parser in libpll is very inefficient on large alignments. e.g for 45 taxa x 6M sites:

• libpll parser: ~1200 sec
• genesis parser: ~3 sec

We should either optimize libpll parser, or use a different one for raxml-ng.

How to cite raxml-ng in publication?

either original version from RAxML/ExaML, or more advanced approach described in this pre-print:

1. http://www.biorxiv.org/content/early/2017/06/10/148536.1
2. https://onlinelibrary.wiley.com/doi/epdf/10.1002/ajb2.1016

When running RAxML-NG on supermuc with --brlen scaled option, I get the following error:

ERROR: ERROR in branch length optimization (LIBPLL-2240): BL opt converged to a worse likelihood score by -779.593183368444443 units

I ran the same search with 11 different seeds and it failed 3 times.
I also happens when I do not use LG4X (but I still have LG4M).
I do not reproduce with the flag --blopt nr_safe.

See <..>/ALL_1.1.aa/jobs/raxml-ng/lg4m_lg4x_scaled_median/runs/unsafe_runs/raxml_seed2001_rand_scaled_nodes32 on supermuc.

Write python scripts to combine raxml-ng searches with the modules for RF distances, consensus trees etc that Diego has implemented.

using terraphast lib here: https://github.com/terraphast/terraphast

Thanks you so much for this amazing re-write, very much looking forward to using it. Is there any chance to implement the ascertainment bias correction options from RAxML?

Needed more frequently than you'd think. Remeber that ExaML bootstrapping bug though !

Hi,

The fixed amino-acid frequencies supplied by models like WAG and LG do not actually sum to 1.0. This doesn't really affect the rate matrix, since it is going to be rescaled to a fixed rate anyway. However, the frequencies are also used to compute the likelihood at the root. The difference is pretty small, about 1.0e-6, but still wrong I think. Perhaps if user-supplied frequencies are different is more than 1% you should error out instead of rescaling?

It would be nice to fix this so that raxml-ng passes the testiphy tests lg/2 and wag/1. Tests with uniform frequencies or user-supplied frequencies that sum to 1 already pass.

Since I guess this is also a libpll issue, let me know if I should file another bug there.

I'm trying to run raxml-ng on comet for a dataset that I seem to see odd behavior when run with the standard raxml.

With both my trimmed and non-trimmed alignment, at some stage in the calculation, the job hits the following error (full output from one of the two jobs is also attached):
raxml-ng-mpi: /home/saladi1/installers/raxml-ng/src/Optimizer.cpp:33: double Optimizer::optimize_model(TreeInfo&, double): Assertion `cur_loglh - new_loglh < -new_loglh * 1e-14' failed.
slurm_raxml-ng.10546635.comet-10-49.out.txt

Do you have any suggestions of how I might get around this issue? Apologies if this is a trivial error/mistake on my part.