Skip to content

Super lightweight and fast mmCIF/PDB/MOL2 file parser into Pandas DataFrames and backwards writer.

License

Notifications You must be signed in to change notification settings

Ruibin-Liu/MolDF

Repository files navigation



MolDF: molecular structure data processing and analysis with DataFrame

Tests Documentation Status Python PyPI version

Many file formats are about different ways of integrating structured data blocks into a single file in that those blocks are related to each other in some way. The PDBx or mmCIF file format organizes structural biology data into categories and each category contains a structured data block which includes several attributes, and each attribute contains the same number of elements within a category. Those characteristics make a PDBx/mmCIF file naturally to be representable as a Python dict of Pandas DataFrames.

Our moldf package primarily parses a PDBx file (mmCIF file: <pdb_id>.cif) into a Python dict with PDBx category names as keys and contents belonging to the category as the corresponding values. Each category content is parsed as a Pandas DataFrame whose columns are the attribute names. On the other hand, we can write a dict of Pandas DataFrame(s) into a PDBx format in which the dict key(s) are used as category names, the DataFrame column names as attribute names, and the DataFrame row(s) as the corresponding record(s).

The old style PDB file format is not very well structured compared to the new PDBx file format. However, we can make moldf support parsing a PDB file (pdb_id.pdb) into a Python dict of Pandas DataFrames similarly, although many 'blocks' need more post processing. As such, the lines starting with ATOM, HETATM, and TER are read into a category named _atom_site which corresponds to the same category in a mmCIF file. And for NMR models, all ATOM, HETATM, and TER lines are read into a single DataFrame but atoms in a NMR model has the same value in the nmr_model column which is determined by the number in the corresponding MODEL line. SEQRES lines are read into a category named _seq_res.

The TRIPOS MOL2 file format is also supported for reading using the same keyword as the PDB and PDBx files about selecting file categories. Currently, the ATOM, BOND, and MOLECULE are supported.

For the PDB and MOL2 formats, if any other categories are required in your workflow, please raise an issue or PR.

Requirements

Only Pandas is required since we need to export Pandas DataFrames. Python versions >= 3.9 in Linux, Windows, and MacOS are tested.

Install

pip install moldf

Usage examples

1. PDBx file

1.1 If you want to read the 3D coordinates for PDB 1vii into a Pandas DataFrame, and you have downloaded the 1vii.cif file to your current working directory ./, you can:

from moldf import read_pdbx
pdbx_file = './1vii.cif'
pdbx = read_pdbx(pdbx_file, category_names=['_atom_site'])
atoms_df = pdbx['_atom_site']
# 'atoms_df' is a Pandas DataFrame containing the '_atom_site' category which has the detailed 3D coordinates for each atom.

1.2. If you want to read the FASTA sequence of 1vii, you can:

from moldf import read_pdbx
pdb_id = '1vii'
pdbx = read_pdbx(pdb_id=pdb_id, category_names=['_entity_poly'])
fasta_df = pdbx['_entity_poly']
fasta = fasta_df['pdbx_seq_one_letter_code_can'].to_list()[0]  # 1vii only has one sequence
# fasta == 'MLSDEDFKAVFGMTRSAFANLPLWKQQNLKKEKGLF'

The file content of 1VII.cif is fetched from RCSB if only pdb_id is provided.

You can also put a Uniprot ID in the pdb_id keyword to get the AlphaFold2 file from alphafold.ebi.ac.uk.

By default, the fetched content will be saved to a file named <pdb_id>.cif under the directory given by pdbx_file_dir (which by default is your current working directory). You can choose not to save the file by setting save_pdbx_file=False in the read_pdbx function calling.

1.3. You can read them simutanously:

from moldf import read_pdbx
pdbx_file = './1vii.cif'
pdbx = read_pdbx(pdbx_file, category_names=['_entity_poly', '_atom_site'])
atoms_df = pdbx['_atom_site']
fasta_df = pdbx['_entity_poly']

Putting a list of category names to category_names, you will get them if they are in the PDBx file.

1.4. You can parse the whole file by using 'all':

from moldf import read_pdbx
pdbx_file = './1vii.cif'
pdbx = read_pdbx(pdbx_file, category_names=['all'])
atoms_df = pdbx['_atom_site']
fasta_df = pdbx['_entity_poly']
# and more

1.5. Write back to a PDBx file:

from moldf import read_pdbx, write_pdbx
pdbx_file = './1vii.cif'
pdbx = read_pdbx(pdbx_file, category_names=['all'])
keep = ['_atom_site', '_entity_poly']  # suppose we only want to keep the FASTA sequence and 3D coordinates.
pdbx_keep = {k: v for k, v in pdbx.items() if k in keep}
write_pdbx(pdbx_keep, '1vii_save.cif')

2. PDB file

2.1. For reading the atomic information in a PDB file 1vii.pdb:

from moldf import read_pdb
pdb_file = './1vii.pdb'
pdb = read_pdb(pdb_file=pdb_file, category_names=['_atom_site'])  # We use '_atom_site' here to mirror the mmCIF format and it is the default
atoms_df = pdb['_atom_site']
# 'atoms_df' is a Pandas DataFrame containing the '_atom_site' category which has the detailed 3D coordinates for each atom.

2.2. Suppose we only want to keep the protein residue atoms in 5u8l.pdb:

from moldf import read_pdb, write_pdb
pdb_file = './5u8l.pdb'
pdb = read_pdb(pdb_file=pdb_file, category_names=['_atom_site'])
df = pdb['_atom_site']
df = df[df.record_name == 'ATOM']
pdb['_atom_site'] = df
write_pdb(pdb, '5u8l_nohetero.pdb')
# The '5u8l_nohetero.pdb' file contains only the protein residues.

The read_pdb function can parse PDB files generated by Chimera by default. You can set allow_chimera=False in its input to fully follow the standard PDB format (although I don't see a use case).

For using a pdb_id or Uniprot ID (as pdb_id) to fetch a RCSB or AlphaFold2 PDB file, it is almost the same as the above read_pdbx example, except that the keywords are save_pdb_file and pdb_file_dir for saving the fetched content.

The write_pdb function can write PDB files that can be parsed by Chimera by setting allow_chimera=True. allow_chimera=False by default so that the output PDB files follow the standard PDB format strictly.

Since our package can read from and write to PDB files containing NMR models, it is straightforward to read and write trajectory files saved as PDB files by molecular dynamics software, if different frames are surrounded by pairs of MODEL and ENDMDL lines.

3. MOL2 file

For example, to read the test.mol2 file in the tests/test_files folder in this repository:

from moldf import read_mol2
mol2_file = './tests/test_files/test.mol2'
mol2 = read_mol2(test_file)
atoms_df = mol2['ATOM']  # The 'ATOM' category as a DataFrame.
bonds_df = mol2['BOND']  # The 'BOND' category as a DataFrame.

Contributing

  1. Fork this repository
  2. Create a branch for your fix of an issue/feature/doc (git checkout -b my-fix)
  3. Stage your changes, run pre-commit, and run pytest until every thing looks fine
  4. Commit your changes (git commit -am 'Added some feature/fix/doc')
  5. Push to the branch (git push origin my-fix)
  6. Create new Pull Request

For code quality control, we use pre-commit hooks for formatting and type hints.

Try your best to add test cases for bug fixes and new features. We use pytest for testing.

For documentation, we should try to follow the Google Python docstrings style.