pdbx2df 0.5.6

A python package to parse PDBx file into Pandas DataFrames.

pdbx2df

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 pdbx2df 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 pdbx2df 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, currently only 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.

Only requirement

• Pandas (>=1.0)

Install

pip install pdbx2df

Usage examples

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 pdbx2df 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.

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

from pdbx2df import read_pdbx
pdbx_file = './1vii.cif'
pdbx = read_pdbx(pdbx_file, 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'

3. You can read them simutanously:

from pdbx2df 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.

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

from pdbx2df 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

5. Write back to a PDBx file:

from pdbx2df 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')

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

from pdbx2df import read_pdb
pdb_file = './1vii.pdb'
pdb = read_pdb(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.

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

from pdbx2df import read_pdb, write_pdb
pdb_file = './5u8l.pdb'
pdb = read_pdb(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).

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.