Problem with partitioned analysis.

[Filipe Romero]

Hello,

I'm in trouble when using IQ-tree for a partitioned analysis with nucleotide sequence data. I think it may be because I have to much partitions, one for each gene, with more than 10000 concatenated loci for most datasets. To evaluate this I made a smaller dataset with 2073 loci and had the same problem. It appears that the estimated time to conclude the analysis only increases along the run, which is unexpected. I'm sending my alignment, the partition file and the command I used. The bug report shown is:

terminate called after throwing an instance of 'std::bad_alloc'
  what():  std::bad_alloc

*** IQ-TREE CRASHES WITH SIGNAL ABORTED
*** For bug report please send to developers:
***    Log file: apes_part_min_entropia.txt.log
***    Alignment files (if possible)
Abortado (imagem do núcleo gravada)

My command is:

iqtree-omp -nt AUTO -s apes_entropia_min_concat.nex -spp apes_part_min_entropia.txt -m TESTMERGE -bb 1000 -alrt 1000 -rcluster 10  -redo

The files:

https://drive.google.com/file/d/0B_pZ60bYrI6eS0dfMG9hR0VGTFU/view?usp=sharing
https://drive.google.com/file/d/0B_pZ60bYrI6eM2U5d3lJNXRKSms/view?usp=sharing
https://drive.google.com/file/d/0B_pZ60bYrI6ecVFIN3BDa1UtYzQ/view?usp=sharing

I wonder if it would be a problem if I made a partition for each set of genes that fits with a model. What are the consequences of this for the inference of topologies and branch lengths?
It seems to me that this has not been fully explored. At least I couldn't find anything about this in the literature.

Thank you in advance,

Filipe.

[Bui Quang Minh]

Hi Filipe,

yes ‘bad_alloc' message means that you don’t have enough RAM for such large data set. Beginning of log file:

Host:    DESKTOP-12KMH6R (AVX2, FMA3, 7 GB RAM)

This is too small RAM. Try another one. You can copy all current output files to another PC so that IQ-TREE can restart from checkpoint to save time.

I wonder if it would be a problem if I made a partition for each set of genes that fits with a model. What are the consequences of this for the inference of topologies and branch lengths?
It seems to me that this has not been fully explored. At least I couldn't find anything about this in the literature.

-m TESTMERGE should do the job, where it tries to merge partitions (see PartitionFinder). So this should reduce the fear of over-parameterization.

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.

--

Bui Quang Minh

Center for Integrative Bioinformatics Vienna (CIBIV)

Campus Vienna Biocenter 5, VBC5, Ebene 1

A-1030 Vienna, Austria

Phone: ++43 1 4277 74326

Email: minh.bui (AT) univie.ac.at

[Filipe Romero]

Thank you, Minh!