Uncaught exception: Failure("Internal edge became -4.62762e-07 when attaching subtrees to

[Jamie Kwok]

Hello, when running guppy as part of PICRUSt2's pipeline, I got the following error after the pipeline finished EPA-ng part:

(...truncated, last lines shown)
INFO     ______ ____   ___           _   __ ______
        / ____// __ \ /   |         / | / // ____/
       / __/  / /_/ // /| | ______ /  |/ // / __
      / /___ / ____// ___ |/_____// /|  // /_/ /
     /_____//_/    /_/  |_|      /_/ |_/ \____/ (v0.2.1-beta)
INFO Number of ranks: 1
INFO Output file: place_seqs_tmp_2026r561/epa_out/epa_result.jplace
INFO 541 Sequences done!
INFO Time spent placing: 11s
INFO Elapsed Time: 14s
guppy tog place_seqs_tmp_2026r561/epa_out/epa_result.jplace -o OUTPUT.tre
Uncaught exception: Failure("Internal edge became -4.62762e-07 when attaching subtrees to 32476")
Fatal error: exception Failure("Internal edge became -4.62762e-07 when attaching subtrees to 32476")
Error running this command:
guppy tog place_seqs_tmp_2026r561/epa_out/epa_result.jplace -o OUTPUT.tre

I also cross-posted this question on the PICRUSt-users Google group here, with some more notes, particularly that this error only occurs for one of my four microbiome samples. Do you have ideas on this? Thank you very much for your attention.

Best regards,
Jamie

[Erick Matsen]

Not sure, but pplacer can behave badly when there are very short edges in your reference tree. If that's the case for you I suggest sparsifying your reference tree.

--
You received this message because you are subscribed to the Google Groups "pplacer users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pplacer-user...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Frederick "Erick" Matsen, Associate Member
Fred Hutchinson Cancer Research Center
http://matsen.fredhutch.org/

[Jamie Kwok]

Hi Erick, turns out I was giving it an incorrect input file, in that file there are multiple identical sequences, which is consistent with your guess. After correcting the input file it runs fine. Thank you very much!

Best regards,

Jamie

To unsubscribe from this group and stop receiving emails from it, send an email to pplacer-users+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--

Frederick "Erick" Matsen, Associate Member
Fred Hutchinson Cancer Research Center
http://matsen.fredhutch.org/

--
You received this message because you are subscribed to a topic in the Google Groups "pplacer users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/pplacer-users/jWWVWLLtb6Q/unsubscribe.
To unsubscribe from this group and all its topics, send an email to pplacer-users+unsubscribe@googlegroups.com.

[Jamie Kwok]

Oh wait nevermind, I mixed up something, the problem isn't solved.

I'll get back to you if I have updates from PICRUSt2 side. Thanks.

Best regards,

Jamie

[Gavin Douglas]

Hi Erick,

I'm also troubleshooting this problem (based on the sample .jplace file). We've narrowed it down to just 2 sequences - I'll post the section of the jplace file describing them below. Does anything strike you as off about the length (or other) values? I'm not sure why a negative internal edge would be created based on the below values, but hopefully something will jump out to you!

Thanks,

Gavin

"placements":

  [

    {"p": [

      [25398, -1842850.461980, 1.000000, 1.618659, 3.412664]

      ],

    "m": {"entropy": 0.000000},

    "n": ["seq_8"]

    },

    {"p": [

      [32476, -1842429.434979, 1.000000, 0.010497, 0.000113]

      ],

    "m": {"entropy": 0.000000},

    "n": ["seq_7"]

    }

  ],

[Erick Matsen]

Oh, you are using an EPA-ng jplace file. I tried hard to make them talk to each other, but in the end the concordance isn't perfect. EPA-ng allows the tree to flex. 

Very sorry for the pain. I'd think that picrust would have had this problem before??

--

You received this message because you are subscribed to the Google Groups "pplacer users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to pplacer-user...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[Gavin Douglas]

Hi Erick,

Thanks for the quick reply - ok that's good to know! I'm working on a new version of PICRUSt which uses phylogenetic placement that isn't released yet (the previous version used closed-reference OTU picking) and I hadn't run into this problem before. It looks like the tool "gappa" written by the authors of the genesis C++ library (which overlaps with the epa-ng team) has a function to convert from jplace to newick so it might be the best option for users using epa-ng-formatted jplace files in the future.

Best,

Gavin