Handling Assemblies in PDB vs PDBX

[zachcp]

I've found myself in a bit of a rabbit hole. When looking into bumping biotite (#732, #733), I noticed that the only blocker for upgrading was due to the use of a few functions in 'OldCIF.py'. In fact, there doesn't seem to be ANY current use in the codebase so I think it should probably be removed (WIP: #733). However, I am running issues in to the tests which DO use oldCIF in a few places within a single file: test_assemblies.py

To address that issue, I optimistically updated the code using OldCIF to use the modern pdbx code. However, the new CIFAssemblyParser had a different API from the previos version as I will show below

import molecularnodes.entities.molecule.pdb as pdb
import molecularnodes.entities.molecule.pdbx as pdbx
import biotite.structure.io.pdb as biotite_pdb
import biotite.structure.io.pdbx as biotite_cif

file_pdb = "tests/data/1f2n.pdb"
file_cif = "tests/data/1f2n.cif"

## PDB
pdb_file = biotite_pdb.PDBFile.read(file_pdb)
atoms = biotite_pdb.get_structure(pdb_file, model=1)
ref_assembly = biotite_pdb.get_assembly(pdb_file, model=1)
test_parser = pdb.PDBAssemblyParser(pdb_file)
assembly_id = test_parser.list_assemblies()[0]
test_transformations_pdb = test_parser.get_transformations(assembly_id)


## PDBX
cif_file = biotite_cif.CIFFile().read(file_cif)
atoms = biotite_cif.get_structure(
    # Make sure `label_asym_id` is used instead of `auth_asym_id`
    cif_file,
    model=1,
    use_author_fields=False,
)
ref_assembly = biotite_cif.get_assembly(cif_file, model=1)
test_parser = pdbx.CIFAssemblyParser(cif_file)
assembly_id = test_parser.list_assemblies()[0]
test_transformations_pdbx  = test_parser.get_transformations(assembly_id)

# test_transformations_pdb
# len(test_transformations_pdb) ==  60
...
...
 (['A', 'B', 'C'],
  [[-0.791048, 0.48869, -0.368003, 156.05355],
   [-0.198289, 0.364253, 0.909946, -51.75154],
   [0.578727, 0.792781, -0.191239, 44.70419],
   [0.0, 0.0, 0.0, 1.0]]),
 (['A', 'B', 'C'],
  [[-0.234445, 0.605831, -0.760266, 144.34027],
   [0.538206, 0.732159, 0.417466, -69.17376],
   [0.809549, -0.311307, -0.497713, 50.25899],
   [0.0, 0.0, 0.0, 1.0]])]

# test_transformations_pdbx
# len(test_transformations_pdbx) ==  60
...
....
 {'chain_ids': ['A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I'],
  'matrix': [[-0.7910475, 0.48868961, -0.36800314, 156.05355],
   [-0.19828949, 0.36425266, 0.90994575, -51.75154],
   [0.57872716, 0.79278147, -0.19123915, 44.70419],
   [2.1643740543e-314,
    2.1643740543e-314,
    2.1643740543e-314,
    2.1643740543e-314]],
  'pdb_model_num': 0},
 {'chain_ids': ['A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I'],
  'matrix': [[-0.23444536, 0.60583125, -0.76026565, 144.34027],
   [0.5382063, 0.7321588, 0.41746554, -69.17376],
   [0.80954886, -0.31130691, -0.49771343, 50.25899],
   [2.1643740543e-314,
    2.1643740543e-314,
    2.1643740543e-314,
    2.1643740543e-314]],
  'pdb_model_num': 0}]

So: I am not an expert in these assemblies but I notice that:

PDB

3 chains.

CIF

3 chains plus chains that are all HETATM / WATER

What to do?

Considering:

• oldcif.py is not used/imported in the codebase
• pdbx via biotite seems to be the way of the future
• the only place old cif is used is in the tests (e.g. but NOT in the actual code )

I think we should:

• nuke oldcif.py
• update the tests.

However:

• we would need to define the correct behavior derived from the discrepancy above.
• after writing this all, however, I think I know the right temporary solution which is just to shim the test to convert the dict-based solution to an iterated list

[zachcp]

Note: I have only a minimal fix in #732 . It turns out OldCIF.py is used extensively via the load_local function.