search

amkozlov / raxml-ng

RAxML Next Generation: faster, easier-to-use and more flexible
GNU Affero General Public License v3.0
385 stars 64 forks source link

Assertion `limit` failed utree_distances #130

Closed padpadpadpad closed 2 years ago

padpadpadpad commented 2 years ago

Hi all

I am using raxml-ng that I installed via conda (RAxML-NG v. 1.0.3) on a ~250bp alignment of 3000 OTUs.

It got quite far and then I got this error when i checked on it this morning.

raxml-ng: /opt/conda/conda-bld/raxml-ng_1626880669841/work/libs/pll-modules/src/tree/utree_distances.c:790: _cmp_splits: Assertion `limit' failed. Aborted (core dumped)

This was after running the command:

raxml-ng --msa alignment_ps_otu_99percent_filtered.fasta --model GTR+G --seed 2 --prefix raxml/myxo1 --threads 4 --outgroup TCGCTGGCCGAATCGGAAGGTCTGACTCCGCAGGAACTGATCAACGCAAAGCCGGTGGCTGCCGCGATCAAGGAATTCTTCGGATCCTCGCAGCTTTCGCAGTTCATGGACCAGAACAATCCGCTTTCGGAAGTCACCCACAAGCGCCGCGTTTCGGCACTTGGACCAGGTGGCCTGACCCGCGAGCGTGCTGGCTTCGAAGTACGC

Attached is my alignment and my log file. Any ideas much appreciated. log_and_alignment.zip

amkozlov commented 2 years ago

Hi @padpadpadpad,

thanks for reporting! There is a bug when trying to root a collapsed multifurcating tree. So until I fix it, the easiest workaround would to be resume the analysis without --outgroup option, and then run a separate raxml-ng --evaluate --tree myxo1.raxml.bestTree --outgroup X,Y if needed.

Side note: apparently, you are trying to infer a "tree of reads", which is generally a bad idea due to insufficient signal (see eg. this discussion). Please consider using phylogenetic placement instead: https://github.com/Pbdas/epa-ng

Side note 2: please don't use sequences as taxon names, this was really inconvenient to work with.

padpadpadpad commented 2 years ago

Hi @amkozlov I shall look into everything you have suggested!

Thank you so much for your quick and in-depth response.

In response to side note 2, I am trying to create the phylogenetic tree from the output of a dada2 run which automatically names the sequences the same as the sequence itself. I will make sure to change this in future for tree making and reporting purposes. Sorry for the inconvenience!

amkozlov commented 2 years ago

fixed in v1.1.0