BUG: IQtree crashes during bootstrap running with predefined codon model

[Karen Siu Ting]

Hi Minh,

I was trying to run a non-parametric bootstrap analysis of a nucleotide alignment using codon models in my server (Linux-64 bit, my node is: Host:    node002 (AVX2, FMA3, 1009 GB RAM)

Kernel:  AVX+FMA - 8 threads (16 CPU cores detected)

Command used is:

iqtree-omp -s gene_untrimmed_280218_QC.fasta -st CODON -m GY+F+R10 -b 100 -nt 8

I had provided the model here because I had tried running it before without specifying the model and it crashed. After several attempts, I got it to the point where it selected the best model, which is that one specified above.

So in the last run, after specifying the model, it run fine for 4 bootstraps, but then it crashed in the 5th. I tried re-submitting the job but always crashes at this point.

I get the following error message:

CHECKPOINT: Initial tree restored

NOTE: 1343 MB RAM (1 GB) is required!

Estimate model parameters (epsilon = 0.100)

1. Initial log-likelihood: -520358.546

ERROR: phylokernelnew.h:2951: double PhyloTree::computeLikelihoodBranchGenericSIMD(PhyloNeighbor*, PhyloNode*) [with VectorClass = Vec4d; bool SAFE

_NUMERIC = true; bool FMA = true; bool SITE_MODEL = false]: Assertion `!std::isnan(tree_lh) && "Numerical underflow for lh-branch"' failed.

ERROR: STACK TRACE FOR DEBUGGING:

ERROR: 1   funcAbort()

ERROR: 3   gsignal()

ERROR: 4   abort()

ERROR: 6   double PhyloTree::computeLikelihoodBranchGenericSIMD<Vec4d, true, true, false>(PhyloNeighbor*, PhyloNode*)

ERROR: 7   PhyloTree::computeLikelihood(double*)

ERROR: 8   PhyloTree::computeFunction(double)

ERROR: 9   Optimization::minimizeOneDimen(double, double, double, double, double*, double*)

ERROR: 10   PhyloTree::optimizeTreeLengthScaling(double, double&, double, double)

ERROR: 11   RateFree::optimizeWithEM()

ERROR: 12   ModelFactory::optimizeParameters(int, bool, double, double)

ERROR: 13   ModelFactory::optimizeParametersGammaInvar(int, bool, double, double)

ERROR: 14   IQTree::optimizeModelParameters[abi:cxx11](bool, double)

ERROR: 15   runTreeReconstruction(Params&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&, IQTree*&, ModelCheckpo

int&)

ERROR: 16   runStandardBootstrap(Params&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&, Alignment*, IQTree*)

ERROR: 17   runPhyloAnalysis(Params&, Checkpoint*)

ERROR: 18   main()

ERROR: 19   __libc_start_main()

ERROR:

ERROR: *** IQ-TREE CRASHES WITH SIGNAL ABORTED

ERROR: *** For bug report please send to developers:

ERROR: ***    Log file: gene_untrimmed_280218_QC.fasta.log

ERROR: ***    Alignment files (if possible)

/cm/local/apps/sge/current/spool/node002/job_scripts/3100568: line 14: 28590 Aborted                 (core dumped) ~/bin/iqtree-omp -s gene_untrimm

ed_280218_QC.fasta -st CODON -m GY+F+R10 -b 100 -nt 8

Is there anything I'm supposed to do differently? or is it a problem of my software? I'm running a version compiled in the actual cluster because pre-compiled versions always crashed in my server.

Any help would be very appreciated!

Thanks!

Karen.

[Karen Siu Ting]

Oh I forgot to add, the seed number for this crashed job was:

===> START BOOTSTRAP REPLICATE NUMBER 5

Creating bootstrap alignment (seed: 268635)...

CHECKPOINT: Initial tree restored

...

[Minh Bui]

Hi Karen, 

Thanks for the detailed report. Do you see any WARNINGs in the log-file? My guess now is that +R10 can be over-fitting for your alignment, causing numerical problem.

But anyway please send me the alignment and all output files via my personal email for debugging purpose.

Cheers

Minh

--
You received this message because you are subscribed to the Google Groups "IQ-TREE" group.
To unsubscribe from this group and stop receiving emails from it, send an email to iqtree+un...@googlegroups.com.
To post to this group, send email to iqt...@googlegroups.com.
Visit this group at https://groups.google.com/group/iqtree.
For more options, visit https://groups.google.com/d/optout.

[Minh Bui]

Hi Karen, 

While fixing this bug, I notice some obvious error in your alignments:

That means, the start codon (ATG) are not correctly aligned. These 3 middle sequences are shifted forward and wrongly aligned, which might cause long branches leading to them and numerical problems as you observed. So I would first go back to correct the alignment again (even manually if the alignment program does not do a good job), before running IQ-TREE.

Cheers

Minh