AF3​​​​​​​ parse函数解读

AlphaFold3 中mmcif_parsing模块的parse函数的主要作用是 解析 mmCIF 文件内容，提取蛋白质结构信息，转换为 MmcifObject，并返回解析结果。

源代码：

@functools.lru_cache(16, typed=False)
def parse(
    *, file_id: str, mmcif_string: str, catch_all_errors: bool = True
) -> ParsingResult:
    """Entry point, parses an mmcif_string.

    Args:
      file_id: A string identifier for this file. Should be unique within the
        collection of files being processed.
      mmcif_string: Contents of an mmCIF file.
      catch_all_errors: If True, all exceptions are caught and error messages are
        returned as part of the ParsingResult. If False exceptions will be allowed
        to propagate.

    Returns:
      A ParsingResult.
    """
    errors = {}
    try:
        parser = PDB.MMCIFParser(QUIET=True)
        handle = io.StringIO(mmcif_string)
        full_structure = parser.get_structure("", handle)
        first_model_structure = _get_first_model(full_structure)
        # Extract the _mmcif_dict from the parser, which contains useful fields not
        # reflected in the Biopython structure.
        parsed_info = parser._mmcif_dict  # pylint:disable=protected-access

        # Ensure all values are lists, even if singletons.
        for key, value in parsed_info.items():
            if not isinstance(value, list):
                parsed_info[key] = [value]

        header = _get_header(parsed_info)

        # Determine the protein chains, and their start numbers according to the
        # internal mmCIF numbering scheme (likely but not guaranteed to be 1).
        valid_chains = _get_protein_chains(parsed_info=parsed_info)
        if not valid_chains:
            return ParsingResult(
                None, {(file_id, ""): "No protein chains found in this file."}
            )
        seq_start_num = {
            chain_id: min([monomer.num for monomer in seq])
            for chain_id, seq in valid_chains.items()
        }

        # Loop over the atoms for which we have coordinates. Populate two mappings:
        # -mmcif_to_author_chain_id (maps internal mmCIF chain ids to chain ids used
        # the authors / Biopython).
        # -seq_to_structure_mappings (maps idx into sequence to ResidueAtPosition).
        mmcif_to_author_chain_id = {}
        seq_to_structure_mappings = {}
        for atom in _get_atom_site_list(parsed_info):
            if atom.model_num != "1":
                # We only process the first model at the moment.
                continue

            mmcif_to_author_chain_id[atom.mmcif_chain_id] = atom.author_chain_id

            if atom.mmcif_chain_id in valid_chains:
                hetflag = " "
                if atom.hetatm_atom == "HETATM":
                    # Water atoms are assigned a special hetflag