Class FDProtein

flowdock/utils/data_utils.py:732–773

Protein structure representation.

@dataclasses.dataclass()
class FDProtein:
    """Protein structure representation."""

    # The first entry stores amino acid sequence in letter representation.
    # The second entry stores a 0-1 mask for observed standard residues.
    # Non-standard residues are mapped to <mask> to interact with protein language models.
    letter_sequences: List[Tuple[str, str, np.ndarray]]

    # Cartesian coordinates of atoms in angstroms. The atom types correspond to
    # residue_constants.atom_types, i.e. the first three are N, CA, CB.
    atom_positions: np.ndarray  # [num_res, atom_type_num, 3]

    # Amino-acid type for each residue represented as an integer between 0 and
    # 20, where 20 is 'X'.
    aatype: np.ndarray  # [num_res]

    # Added
    # Integer for atom type.
    atomtypes: np.ndarray  # [num_res, element_type_num]

    # Binary float mask to indicate presence of a particular atom. 1.0 if an atom
    # is present and 0.0 if not. This should be used for loss masking.
    atom_mask: np.ndarray  # [num_res, atom_type_num]

    # Residue index as used in PDB. It is not necessarily continuous or 0-indexed.
    residue_index: np.ndarray  # [num_res]

    # 0-indexed number corresponding to the chain in the protein that this residue
    # belongs to.
    chain_index: np.ndarray  # [num_res]

    # B-factors, or temperature factors, of each residue (in sq. angstroms units),
    # representing the displacement of the residue from its ground truth mean
    # value.
    b_factors: np.ndarray  # [num_res, atom_type_num]

    def __post_init__(self):
        if len(np.unique(self.chain_index)) > PDB_MAX_CHAINS:
            raise ValueError(
                f"Cannot build an instance with more than {PDB_MAX_CHAINS} chains "
                "because these cannot be written to PDB format."
            )


@beartype