OBBioMatchError

[kuraisle]

Hi,

I'm trying to run a structure through arpeggio. It works fine on the web server. I've used the "clean_pdb.py" script from PDBTools on it and I still get the OBBioMatchError. None of my atoms have alternative locations and all of my atom serial numbers are unique. If there's anything else I could try, that would be great!

Thanks,
James

[hippolytej]

Hey,

I'm having the same issue here, with a structure that works just fine on the web server, and fails with this package.

I am using OpenMM's PDBFixer to clean an input file download from RCSB, attached here: clean_5lxr.pdb.zip

Steps to reproduce:

First download and unzip input file clean_5lxr.pdb.zip, then:

conda create -n arpeggio-env python=3.7
conda activate arpeggio-env
conda install -c openbabel openbabel
pip install git+https://github.com/PDBeurope/arpeggio.git@master#egg=arpeggio
arpeggio clean_5lxr.pdb

Full Traceback

INFO//16:44:50.496//Program begin.
WARNING//16:44:50.496//No selection was perceived. Defaults into full structure!!
/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/Bio/PDB/StructureBuilder.py:92: PDBConstructionWarning: WARNING: Chain A is discontinuous at line 3359.
  PDBConstructionWarning,
/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/Bio/PDB/StructureBuilder.py:92: PDBConstructionWarning: WARNING: Chain B is discontinuous at line 3367.
  PDBConstructionWarning,
/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/Bio/PDB/StructureBuilder.py:92: PDBConstructionWarning: WARNING: Chain A is discontinuous at line 3368.
  PDBConstructionWarning,
/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/Bio/PDB/StructureBuilder.py:92: PDBConstructionWarning: WARNING: Chain B is discontinuous at line 3411.
  PDBConstructionWarning,
DEBUG//16:44:50.536//Loaded PDB structure (BioPython)
DEBUG//16:44:51.450//Loaded PDB structure (OpenBabel)
ERROR//16:44:51.451//An OpenBabel atom could not be matched to a BioPython counterpart.
Traceback (most recent call last):
  File "/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/arpeggio/core/interactions.py", line 1995, in _establish_structure_mappping
    biopython_atom = serial_to_bio[serial]
KeyError: 0

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/bin/arpeggio", line 8, in <module>
    sys.exit(main())
  File "/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/arpeggio/scripts/process_protein_cli.py", line 79, in main
    run_arpeggio(args)
  File "/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/arpeggio/scripts/process_protein_cli.py", line 89, in run_arpeggio
    i_complex = InteractionComplex(args.filename, args.vdw_comp, args.interacting, args.ph)
  File "/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/arpeggio/core/interactions.py", line 92, in __init__
    self._establish_structure_mappping()
  File "/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/arpeggio/core/interactions.py", line 1998, in _establish_structure_mappping
    raise OBBioMatchError(serial)
arpeggio.core.exceptions.OBBioMatchError: 0

Thanks for your help,

Hippolyte

[kuraisle]

Hey,

I'm having the same issue here, with a structure that works just fine on the web server, and fails with this package.

I am using OpenMM's PDBFixer to clean an input file download from RCSB, attached here: clean_5lxr.pdb.zip

Steps to reproduce:

First download and unzip input file clean_5lxr.pdb.zip, then:

conda create -n arpeggio-env python=3.7
conda activate arpeggio-env
conda install -c openbabel openbabel
pip install git+https://github.com/PDBeurope/arpeggio.git@master#egg=arpeggio
arpeggio clean_5lxr.pdb

Full Traceback

INFO//16:44:50.496//Program begin.
WARNING//16:44:50.496//No selection was perceived. Defaults into full structure!!
/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/Bio/PDB/StructureBuilder.py:92: PDBConstructionWarning: WARNING: Chain A is discontinuous at line 3359.
  PDBConstructionWarning,
/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/Bio/PDB/StructureBuilder.py:92: PDBConstructionWarning: WARNING: Chain B is discontinuous at line 3367.
  PDBConstructionWarning,
/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/Bio/PDB/StructureBuilder.py:92: PDBConstructionWarning: WARNING: Chain A is discontinuous at line 3368.
  PDBConstructionWarning,
/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/Bio/PDB/StructureBuilder.py:92: PDBConstructionWarning: WARNING: Chain B is discontinuous at line 3411.
  PDBConstructionWarning,
DEBUG//16:44:50.536//Loaded PDB structure (BioPython)
DEBUG//16:44:51.450//Loaded PDB structure (OpenBabel)
ERROR//16:44:51.451//An OpenBabel atom could not be matched to a BioPython counterpart.
Traceback (most recent call last):
  File "/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/arpeggio/core/interactions.py", line 1995, in _establish_structure_mappping
    biopython_atom = serial_to_bio[serial]
KeyError: 0

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/bin/arpeggio", line 8, in <module>
    sys.exit(main())
  File "/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/arpeggio/scripts/process_protein_cli.py", line 79, in main
    run_arpeggio(args)
  File "/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/arpeggio/scripts/process_protein_cli.py", line 89, in run_arpeggio
    i_complex = InteractionComplex(args.filename, args.vdw_comp, args.interacting, args.ph)
  File "/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/arpeggio/core/interactions.py", line 92, in __init__
    self._establish_structure_mappping()
  File "/Users/hippolyte/opt/miniconda3/envs/arpeggio-env/lib/python3.7/site-packages/arpeggio/core/interactions.py", line 1998, in _establish_structure_mappping
    raise OBBioMatchError(serial)
arpeggio.core.exceptions.OBBioMatchError: 0

Thanks for your help,

Hippolyte

Hi Hippolyte,

Strangely I have since run the python 2.7 version of this and it seems to work (with some warnings). I still have to clean up my structures before using them, but I get a result!

[hippolytej]

Hi @kuraisle thanks for sharing, I've been as well using the py2.7 version, which works just fine with the structures that I have :)
Still, a python 3 version would be much more practical...

[reneeotten]

This is indeed a problem for ever PDB file (downloading the same CIF file from RCSB works). The issue is that the the Id for first residue yields 1 for an MMCIF file, whereas it gives 0 for a PDB file.

import os

import openbabel as ob

from arpeggio.core import protein_reader


PDBid = '2ltq'
basename = os.path.join('arpeggio/tests/test_data/structures', f'{PDBid}')

ob_conv = ob.OBConversion()
ob_conv.SetInFormat('pdb')
mol = ob.OBMol()
ob_conv.ReadFile(mol, f'{basename}.pdb')
pdb_atom_iter = ob.OBMolAtomIter(mol)

mmcif = protein_reader.read_mmcif_to_openbabel(f'{basename}.cif')
mmcif_atom_iter = ob.OBMolAtomIter(mmcif)

pdb_atom = pdb_atom_iter.next()
mmcif_atom = mmcif_atom_iter.next()

print('\n*** ID for the first atom when reading in the PDB or CIF file for '
      f'the same structure ({PDBid}) ***')
print(f'pdb_atom.GetId = {pdb_atom.GetId()}')
print(f'mmcif_atom.GetId = {mmcif_atom.GetId()}')

results in:

*** ID for the first atom when reading in the PDB or CIF file for the same structure (2ltq) ***
pdb_atom.GetId = 0
mmcif_atom.GetId = 1

It seems to me that for the PDB file it is using "standard" OpenBabel functions for reading in the file, whereas an mmCIF file uses in-house written code. I haven't gone through the code in detail where the difference originates from, but this definitely needs to be fixed as running a PDB or mmCIF file should give the same result.

[skelm]

Hi all,

I have a somewhat dirty fix. My _establish_structure_mappping() method in arpeggio/core/interactions.py (starting at line 1981) now looks like this:

def _establish_structure_mappping(self):
    """Maps biopython atoms to openbabel ones and vice-versa.

    Raises:
        OBBioMatchError: If we cant match an OB atom to a biopython
    """
    # FIRST MAP PDB SERIAL NUMBERS TO BIOPYTHON ATOMS FOR SPEED LATER
    # THIS AVOIDS LOOPING THROUGH `s_atoms` MANY TIMES
    serial_to_bio = {x.serial_number: x for x in self.s_atoms}

    # `Id` IS A UNIQUE AND STABLE ID IN OPENBABEL
    # CAN RECOVER THE ATOM WITH `mol.GetAtomById(id)`
    serial_to_ob = {x.GetId(): x for x in ob.OBMolAtomIter(self.ob_mol)}

    if 0 in serial_to_ob and 0 not in serial_to_bio:
        offset = 1
        logging.debug('OB atom serial numbers start with 0, but there are no BioPython atoms with serial number 0. Adding 1 to all OB serials for mapping purposes.')
    else:
        offset = 0

    for serial, ob_atom in serial_to_ob.items():
        # MATCH TO THE BIOPYTHON ATOM BY SERIAL NUMBER
        try:
            biopython_atom = serial_to_bio[serial + offset]
        except KeyError:
            raise OBBioMatchError("Failed to match OB atom to a BioPython atom: {},{} : {}({})".format(serial, serial + offset, ob_atom.GetType()[:1], ob_atom.GetType()))
        
        # Sanity check that elements match, for debugging
        if ob_atom.GetType()[:1] != biopython_atom.element[:1]:
            raise OBBioMatchError("Failed to correctly match atoms with different elements: {},{} : {}({}) != {}({}, {})".format(serial, serial + offset, ob_atom.GetType()[:1], ob_atom.GetType(), biopython_atom.element[:1], biopython_atom.element, biopython_atom.name))

        self.ob_to_bio[ob_atom.GetId()] = biopython_atom
        self.bio_to_ob[biopython_atom] = ob_atom.GetId()

    logging.debug('Mapped OB to BioPython atoms and vice-versa.')

I'm happy to let the author fix it in a cleaner way, or let someone else do a proper pull request.