Partition merging .best_model format issues

[zivlie]

Hello,
I just encountered unexpected behavior in the output of a partition merging run. In other cases, the same input arguments have produced a .best_model.nex file that works fine as input for tree inference. In this case, the best_model.nex (and best_scheme.nex) caused errors in two ways:

• The CharPartition caused a failure: Reading partition model file doli-100.3t-95p.charsets.best_model.nex ... ERROR: ERROR: Error found at line 118, column 107 (file position 50073): ERROR: CharSet uce not found
  (Line 118 being the the first model and parameters and the first charset after 'charpartition mymodels=', column 107 being the name of the first charset GTR{1.03847,7.30405,1.18463,0.909962,6.12252}+F{0.281908,0.219077,0.219095,0.27992}+G4{0.496605}: uce-10.nexus_uce-10041.nexus_uce-10222.nexus. . . .

• The actual charsets were formatted with a comma before the positions, e.g. charset uce-10100.nexus_uce-10908.nexus_uce-11150.nexus_uce-12129.nexus = , 41250-42026 353378-354529 450784-451389 840072-840900;

If I manually remove the charpartition section, and remove the extra commas and whitespace, then it runs just fine. However, I would like to know how to avoid this in future.
I've attached the log for the merging run and the .best_model.nex file. The latter is renamed as a .txt file to allow upload here.
This was using version 2.3.1, but I had the same errors when trying to do the inference step on version 2.0.3
doli-100.3t-95p.charsets.best_model.txt
doli-100.3t-95p.charsets.log

[roblanf]

Hi @zivlie,

Thanks for pointing out the problem! Could you please try running the same analyses (one that works, and the one above that has issues) on the latest version of IQ-TREE?

You can get the latest version here: https://iqtree.github.io/ (currently 3.0.1).

If you can do that and post the results you get (good or bad!) that will be very useful.

If the bug is still there in 3.0.1, it is very helpful if you are able to also share the alignment, or a if you are also able to make a minimal example (e.g. a toy alignment that displays the problem - you can delete taxon names, and as many taxa and sites as possible). That way we can know when we've fixed it.

Thanks,

Rob

[roblanf]

Yes - I expect that manually editing the charsets should be fine.

My suggestion would be to:

1. Manually edit the partition files and check that they are sensible (e.g. that every original partition and its sites are included; I say this because the bug could be more than just an extra comma)

2. Run the analysis with -p for the partition file and then -m MFP. This will consider all possible models on the final partitioning scheme, which is likely to improve it a lot compared to just using GTR+G on every partition.

Rob

[zivlie]

While I haven't been able to fully test this on 3.0.1 yet, I did have a realization: it's not that a comma is being added out of nowhere, but that "DNA" is missing before the comma.
I don't know if this makes a difference - but perhaps it would if there is a combination of DNA and DNAF partitions?

[DiegoJ-Elias]

Hi, I apologize for jumping into this discussion. I have run into the same issues when estimating gene and site concordance vectors. I have manually modified the charset file to just have the following format " ... charset uce_1 = 1-952;... '''
The analysis initialize but now I am getting the following error : "Reading alignment file 1-952 ... ERROR: File not found 1-952." The only work around I have found is passing the alignment using the flag -s when calculating the gene and site concordance vectors.
I run into this error running the analysis in Iqtree2 and the last version of Iqtree3.

Thank you for your time!
Diego

[roblanf]

Thanks for the info. If anyone can post a reproducible example we will take a look ASAP.

[DiegoJ-Elias]

Hi Rob,

Below are the reduced alignment, the original, and manually edited charsets.
Thank you so much for your time and help!

Best,
Diego

manually_edited_charset-raxml.charsets.txt
Original_charset-raxml.charsets.txt
reduced_aling_IQTREE_error_phylip.txt