Metadata Lost When Loading, Modifying and Writing Structure

[katkanaz]

• gemmi version 0.7.1

Hello, I am trying to modify a structure, so that it only contains sugar residues with either altloc A or B -- in the end I want to create two files, each with the given altloc.
I can traverse the loaded structure and remove the desired atoms, but when writing it, the resulting file does not contain some metadata which I need (e.g. _entity.src_method).

import gemmi
from pathlib import Path

def write_structure(input_structure: Path, output_path: Path) -> None:
    structure = gemmi.read_structure(str(input_structure))
    structure.setup_entities()

    to_remove = []
    
    for model_idx, model in enumerate(structure):
        for chain_idx, chain in enumerate(model):
            for residue_idx, residue in enumerate(chain):
                # simplified example of removing residues/atoms
                # in my code I filter by altloc value on the level of atoms
                if residue.name == "GLC":
                    to_remove.append([model_idx,chain_idx,residue_idx])

    for rm in reversed(to_remove):
        del structure[rm[0]][rm[1]][rm[2]]

    options = gemmi.cif.WriteOptions()
    options.misuse_hash = True
    options.align_pairs = 48
    options.align_loops = 20

    structure.make_mmcif_document().write_file(str(output_path), options)

Looking through the documentation and issues (#60 (comment)), I found it is possible to load the file as a Document and update some parts of it from the modified structure object.

def write_doc(input_structure: Path, output_path: Path) -> None:
    structure = gemmi.read_structure(str(input_structure))
    structure.setup_entities()

    to_remove = []

    for model_idx, model in enumerate(structure):
        for chain_idx, chain in enumerate(model):
            for residue_idx, residue in enumerate(chain):
                # simplified example of removing residues/atoms
                # in my code I filter by altloc value on the level of atoms
                if residue.name == "GLC":
                    to_remove.append([model_idx,chain_idx,residue_idx])

    for rm in reversed(to_remove):
        del structure[rm[0]][rm[1]][rm[2]]


    groups = gemmi.MmcifOutputGroups(True)
    groups.atoms = True

    doc = gemmi.cif.read(str(input_structure))

    # block = doc.find_block(structure.info["_entry.id"])
    block = doc.sole_block()
    structure.update_mmcif_block(block, groups)

    options = gemmi.cif.WriteOptions()
    options.misuse_hash = True
    options.align_pairs = 48
    options.align_loops = 20

    doc.write_file(str(output_path), options)

I would expect, that setting groups.atoms = True would overwrite the ATOM/HETATM lines in the document from the modified structure object. But the resulting file does not have these lines at all.
Setting groups = gemmi.MmcifOutputGroups(True) does include the modified lines in the output, but the metadata I need is lost.

• am I understanding update_mmcif_block() correctly and is there another groups option I need to set?
• if I were able to create a document with updated atoms and retained metadata, would this be a valid file? (can gemmi update the rest of the file to reflect the deletion of some residues?)
• am I overcomplicating it and is there a simpler way to remove all A (or all B) altlocs from sugar residues?

[wojdyr]

There was a somewhat related question last month: #362

[katkanaz]

Thank you for your response. Setting the chem_comp and entity to False solved the problem with the missing metadata. However I have now encountered a different problem. I am using a tool that parses mmCIF files that requires _atom_site.auth_atom_id and _atom_site.auth_comp_id. I read in the documentation that gemmi only reads one of these (auth if present, label otherwise). From what I have seen in the resulting files which were modified as a Structure and written by gemmi, only the label_comp_id and label_atom_id columns are written, with auth_comp_id and auth_atom_id missing. Is there a way to also write the two auth columns back?

I also wanted to ask if Gemmi checks and/or handles a difference between label and auth names/number, if such difference occurs?

Does reading a file as a Structure also handle modifying dependencies, when deleting residuus? Or would it be the same as removing the corresponding atom rows "manually"?

[wojdyr]

There is an option auth_all=True for writing both label_comp_id and auth_comp_id, etc.

Re label/auth differences, only chain names and sequence numbers differ and gemmi handles this.

[wojdyr]

Dependendencies – I'm not sure, I'd need to check. Do you have something specific in mind?

[katkanaz]

There is an option auth_all=True for writing both label_comp_id and auth_comp_id, etc.

Re label/auth differences, only chain names and sequence numbers differ and gemmi handles this.

Thank you so much, that helped. It works well now.

Dependendencies – I'm not sure, I'd need to check. Do you have something specific in mind?

I'm relatively new to mmCIF files so I'm not 100% sure I am using the correct terminology. I was thinking of links as in connectivity relationships between atoms, such as the information in _struct_conn, _chem_comp_bond or similar.