Too many iterations in tqli and numerical problems

[gusta10]

Hi there,

I wonder if I should worry about my results if I got these warnings:

WARNING: States(s) GGC, TGA, TGC, TGG rarely appear in partition gene2 and may cause numerical problems   ... this appears in many other partitions

WARNING: Too many iterations in tqli    ... this keep appearing in my log file, yet the job has not finished

What does 'tqli' means and what are the implications of these warnings in my output tree?

Thanks in advance!

[Minh Bui]

Hi,

These two WARNINGs are related. The one about tqli means that numerical optimisation does not converge, so you may have sub-optimal estimates of model parameters. I believe the beta version 1.7 solved some of these problems, which more often occur with codon models. Therefore, can you please test the beta version (https://github.com/Cibiv/IQ-TREE/releases; scroll down to the current 1.7-beta7) and let me know if this warning disappears or not?

Thanks,

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.

[gusta10]

Hi Minh,

Thank you so much for your quick reply. 

I have tried to install the bete7 version, but I am having an error:

CMake Error at CMakeLists.txt:176 (message):

  GCC version must be at least 5.4!

However, the cluster I work has gcc v6.2.0. 

Any solution you recommend? 

Thanks

[Bui Quang Minh]

Please check gcc installation, I believe the error msg is correct. Alternatively, why don’t you just download the binary?

Minh 

[gusta10]

Thank you Minh,

I was able to run again it iqtree v1.7 - beta7 my alignment but have this error now:

Initial log-likelihood: -1570365.783

Current log-likelihood at step 1: -1497252.607

ERROR: modelmarkov.cpp:1250: virtual void ModelMarkov::decomposeRateMatrix(): Assertion `eigensolver.info() == Eigen::Success' failed.

ERROR: 

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

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

[Minh Bui]

This error message is actually much more informative: the eigenvalues cannot be obtained for the rate matrix (for certain combination of model parameters). I suspect it is due to sparse codon rate matrix. But can you pls send me the input file via my personal email for further inspection?

Thanks, 

Minh

[Minh Bui]

Hi,

I believe this assertion is now fixed in 1.7-beta8 version. Can you please check it and let me know if that’s the case:

https://github.com/Cibiv/IQ-TREE/releases

Thank you!

Minh