ruibin-liu / moldf Goto Github PK

Currently, the priority is to make sure the main function read_pdbx works for all the PDBx files in RCSB and SIFTS.

Further enhancements will be:

1. Write back to PDBx and PDB file format. The main use case is to export key structural data like cartesian coordinates, ligands, and sequence info back to standard PDBx or PDB file format so that other programs can process.
2. Useful scripts/functions to select chains, residues, and atoms either based on Amber or VMD selecting language.

If the current version 0.4.0 passes daily use tests, I will publish it as the 1.0.0 version.

I had an idea of exporting the dict of DataFrames data structure as a joint CSV file (.jcsv).

It should be super easy to implement two functions as write_jcsv and read_jcsv.

xyz file format is quite commonly used in quantum chemistry calculations and many molecule visualization programs like Avogadro.

The read_xyz function should take care of the case with multiple structures. I suggest use model_id as the column name.

It will be better if the comment line can be parsed as well. Check this page for hints to parse that line. Grouping this line and the number of atoms into a new DataFrame should be okay. model_id should be used as well if multiple structures are in the file.

Currently, the read_mol2 does not take care of the case when there are multiple structures.

I have encountered an issue when reading in a protein PDB file where whitespace is not effectively removed.

Source: Q05655

Using the following code:

pdb = read_pdb(pdb_file="AF-Q05655-F1-model_v2.pdb", category_names=['_atom_site'])  # We use '_atom_site' here to mirror the mmCIF format and it is the default
atoms_df = pdb['_atom_site']

# Get values for residue_name
list(atoms_df.residue_name.unique())

This yields:

['MET ',
 'ALA ',
 'PRO ',
 'PHE ',
 'LEU ',
 'ARG ',
 'ILE ',
 'ASN ',
 'SER ',
 'TYR ',
 'GLU ',
 'GLY ',
 'GLN ',
 'ASP ',
 'CYS ',
 'VAL ',
 'LYS ',
 'THR ',
 'TRP ',
 'HIS ']

This whitespace should be trimmed so that filtering can take place properly.

Happy to submit a PR for this.

For NMR structures, there are usually multiple models wrapped between MODEL and ENDMDL records.

Two ways to store several models in a dataframe:

• Just put MODEL and ENDMDL lines as rows like those TER lines; simple approach but not so friendly to downstream data processing except for writing back to a PDB file.
• Add a column named model_index to indicate different models; a little more effort but good for downstream data processing.

The second will be implemented in the next version.