biotite 1.0.1__cp310-cp310-win_amd64.whl → 1.2.0__cp310-cp310-win_amd64.whl

biotite/structure/io/pdbx/convert.py

@@ -3,7 +3,7 @@
 # information.
 
 __name__ = "biotite.structure.io.pdbx"
-__author__ = "Fabrice Allain, Patrick Kunzmann"
+__author__ = "Fabrice Allain, Patrick Kunzmann, Cheyenne Ziegler"
 __all__ = [
     "get_sequence",
     "get_model_count",
@@ -13,6 +13,7 @@ __all__ = [
     "set_component",
     "list_assemblies",
     "get_assembly",
+    "get_sse",
 ]
 
 import itertools
@@ -24,6 +25,10 @@ from biotite.structure.atoms import AtomArray, AtomArrayStack, repeat
 from biotite.structure.bonds import BondList, BondType, connect_via_residue_names
 from biotite.structure.box import unitcell_from_vectors, vectors_from_unitcell
 from biotite.structure.error import BadStructureError
+from biotite.structure.filter import _canonical_aa_list as canonical_aa_list
+from biotite.structure.filter import (
+    _canonical_nucleotide_list as canonical_nucleotide_list,
+)
 from biotite.structure.filter import (
     filter_first_altloc,
     filter_highest_occupancy_altloc,
@@ -36,32 +41,38 @@ from biotite.structure.io.pdbx.bcif import (
 from biotite.structure.io.pdbx.cif import CIFBlock, CIFFile
 from biotite.structure.io.pdbx.component import MaskValue
 from biotite.structure.io.pdbx.encoding import StringArrayEncoding
-from biotite.structure.residues import get_residue_count, get_residue_starts_for
+from biotite.structure.residues import (
+    get_residue_count,
+    get_residue_positions,
+    get_residue_starts_for,
+)
 from biotite.structure.util import matrix_rotate
 
-# Cond types in `struct_conn` category that refer to covalent bonds
-PDBX_COVALENT_TYPES = [
-    "covale",
-    "covale_base",
-    "covale_phosphate",
-    "covale_sugar",
-    "disulf",
-    "modres",
-    "modres_link",
-    "metalc",
-]
-# Map 'struct_conn' bond orders to 'BondType'...
-PDBX_BOND_ORDER_TO_TYPE = {
-    "": BondType.ANY,
-    "sing": BondType.SINGLE,
-    "doub": BondType.DOUBLE,
-    "trip": BondType.TRIPLE,
-    "quad": BondType.QUADRUPLE,
+# Bond types in `struct_conn` category that refer to covalent bonds
+PDBX_BOND_TYPE_ID_TO_TYPE = {
+    # Although a covalent bond, could in theory have a higher bond order,
+    # practically inter-residue bonds are always single
+    "covale": BondType.SINGLE,
+    "covale_base": BondType.SINGLE,
+    "covale_phosphate": BondType.SINGLE,
+    "covale_sugar": BondType.SINGLE,
+    "disulf": BondType.SINGLE,
+    "modres": BondType.SINGLE,
+    "modres_link": BondType.SINGLE,
+    "metalc": BondType.COORDINATION,
+}
+PDBX_BOND_TYPE_TO_TYPE_ID = {
+    BondType.ANY: "covale",
+    BondType.SINGLE: "covale",
+    BondType.DOUBLE: "covale",
+    BondType.TRIPLE: "covale",
+    BondType.QUADRUPLE: "covale",
+    BondType.AROMATIC_SINGLE: "covale",
+    BondType.AROMATIC_DOUBLE: "covale",
+    BondType.AROMATIC_TRIPLE: "covale",
+    BondType.COORDINATION: "metalc",
 }
-# ...and vice versa
 PDBX_BOND_TYPE_TO_ORDER = {
-    # 'ANY' is masked later, it is merely added here to avoid a KeyError
-    BondType.ANY: "",
     BondType.SINGLE: "sing",
     BondType.DOUBLE: "doub",
     BondType.TRIPLE: "trip",
@@ -69,6 +80,10 @@ PDBX_BOND_TYPE_TO_ORDER = {
     BondType.AROMATIC_SINGLE: "sing",
     BondType.AROMATIC_DOUBLE: "doub",
     BondType.AROMATIC_TRIPLE: "trip",
+    # These are masked later, it is merely added here to avoid a KeyError
+    BondType.ANY: "",
+    BondType.AROMATIC: "",
+    BondType.COORDINATION: "",
 }
 # Map 'chem_comp_bond' bond orders and aromaticity to 'BondType'...
 COMP_BOND_ORDER_TO_TYPE = {
@@ -79,11 +94,19 @@ COMP_BOND_ORDER_TO_TYPE = {
     ("SING", "Y"): BondType.AROMATIC_SINGLE,
     ("DOUB", "Y"): BondType.AROMATIC_DOUBLE,
     ("TRIP", "Y"): BondType.AROMATIC_TRIPLE,
+    ("AROM", "Y"): BondType.AROMATIC,
 }
 # ...and vice versa
 COMP_BOND_TYPE_TO_ORDER = {
     bond_type: order for order, bond_type in COMP_BOND_ORDER_TO_TYPE.items()
 }
+CANONICAL_RESIDUE_LIST = canonical_aa_list + canonical_nucleotide_list
+# it was observed that when the number or rows in `atom_site` and `struct_conn`
+# exceed a certain threshold,
+# a dictionary approach is less computation and memory intensive than the dense
+# vectorized approach.
+# https://github.com/biotite-dev/biotite/pull/765#issuecomment-2708867357
+FIND_MATCHES_SWITCH_THRESHOLD = 4000000
 
 _proteinseq_type_list = ["polypeptide(D)", "polypeptide(L)"]
 _nucleotideseq_type_list = [
@@ -146,8 +169,8 @@ def get_sequence(pdbx_file, data_block=None):
     -------
     sequence_dict : Dictionary of Sequences
         Dictionary keys are derived from ``entity_poly.pdbx_strand_id``
-        (often equivalent to chain_id and atom_site.auth_asym_id
-        in most cases). Dictionary values are sequences.
+        (equivalent to ``atom_site.auth_asym_id``).
+        Dictionary values are sequences.
 
     Notes
     -----
@@ -203,9 +226,7 @@ def get_model_count(pdbx_file, data_block=None):
         The number of models.
     """
     block = _get_block(pdbx_file, data_block)
-    return len(
-        _get_model_starts(block["atom_site"]["pdbx_PDB_model_num"].as_array(np.int32))
-    )
+    return len(np.unique((block["atom_site"]["pdbx_PDB_model_num"].as_array(np.int32))))
 
 
 def get_structure(
@@ -296,7 +317,6 @@ def get_structure(
     >>> arr = get_structure(file, model=1)
     >>> print(len(arr))
     304
-
     """
     block = _get_block(pdbx_file, data_block)
 
@@ -307,13 +327,12 @@ def get_structure(
         raise InvalidFileError("Missing 'atom_site' category in file")
 
     models = atom_site["pdbx_PDB_model_num"].as_array(np.int32)
-    model_starts = _get_model_starts(models)
-    model_count = len(model_starts)
+    model_count = len(np.unique(models))
     atom_count = len(models)
 
     if model is None:
         # For a stack, the annotations are derived from the first model
-        model_atom_site = _filter_model(atom_site, model_starts, 1)
+        model_atom_site = _filter_model(atom_site, 1)
         # Any field of the category would work here to get the length
         model_length = model_atom_site.row_count
         atoms = AtomArrayStack(model_count, model_length)
@@ -359,7 +378,7 @@ def get_structure(
                 f"the given model {model} does not exist"
             )
 
-        model_atom_site = _filter_model(atom_site, model_starts, model)
+        model_atom_site = _filter_model(atom_site, model)
         # Any field of the category would work here to get the length
         model_length = model_atom_site.row_count
         atoms = AtomArray(model_length)
@@ -475,16 +494,53 @@ def _fill_annotations(array, atom_site, extra_fields, use_author_fields):
     array.set_annotation("element", atom_site["type_symbol"].as_array(str))
 
     if "atom_id" in extra_fields:
-        array.set_annotation("atom_id", atom_site["id"].as_array(int))
+        if "id" in atom_site:
+            array.set_annotation("atom_id", atom_site["id"].as_array(int))
+        else:
+            warnings.warn(
+                "Missing 'id' in 'atom_site' category. 'atom_id' generated automatically.",
+                UserWarning,
+            )
+            array.set_annotation("atom_id", np.arange(array.array_length()))
         extra_fields.remove("atom_id")
     if "b_factor" in extra_fields:
-        array.set_annotation("b_factor", atom_site["B_iso_or_equiv"].as_array(float))
+        if "B_iso_or_equiv" in atom_site:
+            array.set_annotation(
+                "b_factor", atom_site["B_iso_or_equiv"].as_array(float)
+            )
+        else:
+            warnings.warn(
+                "Missing 'B_iso_or_equiv' in 'atom_site' category. 'b_factor' will be set to `nan`.",
+                UserWarning,
+            )
+            array.set_annotation("b_factor", np.full(array.array_length(), np.nan))
         extra_fields.remove("b_factor")
     if "occupancy" in extra_fields:
-        array.set_annotation("occupancy", atom_site["occupancy"].as_array(float))
+        if "occupancy" in atom_site:
+            array.set_annotation("occupancy", atom_site["occupancy"].as_array(float))
+        else:
+            warnings.warn(
+                "Missing 'occupancy' in 'atom_site' category. 'occupancy' will be assumed to be 1.0",
+                UserWarning,
+            )
+            array.set_annotation(
+                "occupancy", np.ones(array.array_length(), dtype=float)
+            )
         extra_fields.remove("occupancy")
     if "charge" in extra_fields:
-        array.set_annotation("charge", atom_site["pdbx_formal_charge"].as_array(int, 0))
+        if "pdbx_formal_charge" in atom_site:
+            array.set_annotation(
+                "charge",
+                atom_site["pdbx_formal_charge"].as_array(
+                    int, 0
+                ),  # masked values are set to 0
+            )
+        else:
+            warnings.warn(
+                "Missing 'pdbx_formal_charge' in 'atom_site' category. 'charge' will be set to 0",
+                UserWarning,
+            )
+            array.set_annotation("charge", np.zeros(array.array_length(), dtype=int))
         extra_fields.remove("charge")
 
     # Handle all remaining custom fields
@@ -536,7 +592,8 @@ def _parse_inter_residue_bonds(atom_site, struct_conn):
     ]
 
     covale_mask = np.isin(
-        struct_conn["conn_type_id"].as_array(str), PDBX_COVALENT_TYPES
+        struct_conn["conn_type_id"].as_array(str),
+        list(PDBX_BOND_TYPE_ID_TO_TYPE.keys()),
     )
     if "ptnr1_symmetry" in struct_conn:
         covale_mask &= struct_conn["ptnr1_symmetry"].as_array(str, IDENTITY) == IDENTITY
@@ -576,13 +633,14 @@ def _parse_inter_residue_bonds(atom_site, struct_conn):
     atoms_indices_1 = atoms_indices_1[mapping_exists_mask]
     atoms_indices_2 = atoms_indices_2[mapping_exists_mask]
 
-    # Interpret missing values as ANY bonds
-    bond_order = struct_conn["pdbx_value_order"].as_array(str, "")
+    bond_type_id = struct_conn["conn_type_id"].as_array()
     # Consecutively apply the same masks as applied to the atom indices
     # Logical combination does not work here,
     # as the second mask was created based on already filtered data
-    bond_order = bond_order[covale_mask][mapping_exists_mask]
-    bond_types = [PDBX_BOND_ORDER_TO_TYPE[order] for order in bond_order]
+    bond_type_id = bond_type_id[covale_mask][mapping_exists_mask]
+    # The type ID is always present in the dictionary,
+    # as it was used to filter the applicable bonds
+    bond_types = [PDBX_BOND_TYPE_ID_TO_TYPE[type_id] for type_id in bond_type_id]
 
     return BondList(
         atom_site.row_count,
@@ -593,9 +651,20 @@ def _parse_inter_residue_bonds(atom_site, struct_conn):
 def _find_matches(query_arrays, reference_arrays):
     """
     For each index in the `query_arrays` find the indices in the
-    `reference_arrays` where all query values the reference counterpart.
+    `reference_arrays` where all query values match the reference counterpart.
     If no match is found for a query, the corresponding index is -1.
     """
+    if (
+        query_arrays[0].shape[0] * reference_arrays[0].shape[0]
+        <= FIND_MATCHES_SWITCH_THRESHOLD
+    ):
+        match_indices = _find_matches_by_dense_array(query_arrays, reference_arrays)
+    else:
+        match_indices = _find_matches_by_dict(query_arrays, reference_arrays)
+    return match_indices
+
+
+def _find_matches_by_dense_array(query_arrays, reference_arrays):
     match_masks_for_all_columns = np.stack(
         [
             query[:, np.newaxis] == reference[np.newaxis, :]
@@ -623,6 +692,38 @@ def _find_matches(query_arrays, reference_arrays):
     return match_indices
 
 
+def _find_matches_by_dict(query_arrays, reference_arrays):
+    # Convert reference arrays to a dictionary for O(1) lookups
+    reference_dict = {}
+    ambiguous_keys = set()
+    for ref_idx, ref_row in enumerate(zip(*reference_arrays)):
+        ref_key = tuple(ref_row)
+        if ref_key in reference_dict:
+            ambiguous_keys.add(ref_key)
+            continue
+        reference_dict[ref_key] = ref_idx
+
+    match_indices = []
+    for query_idx, query_row in enumerate(zip(*query_arrays)):
+        query_key = tuple(query_row)
+        occurrence = reference_dict.get(query_key)
+
+        if occurrence is None:
+            # -1 indicates that no match was found in the reference
+            match_indices.append(-1)
+        elif query_key in ambiguous_keys:
+            # The query cannot be uniquely matched to an atom in the reference
+            raise InvalidFileError(
+                f"The covalent bond in the 'struct_conn' category at index "
+                f"{query_idx} cannot be unambiguously assigned to atoms in "
+                f"the 'atom_site' category"
+            )
+        else:
+            match_indices.append(occurrence)
+
+    return np.array(match_indices)
+
+
 def _get_struct_conn_col_name(col_name, partner):
     """
     For a column name in ``atom_site`` get the corresponding column name
@@ -661,21 +762,26 @@ def _filter_altloc(array, atom_site, altloc):
         raise ValueError(f"'{altloc}' is not a valid 'altloc' option")
 
 
-def _get_model_starts(model_array):
-    """
-    Get the start index for each model in the arrays of the
-    ``atom_site`` category.
-    """
-    _, indices = np.unique(model_array, return_index=True)
-    indices.sort()
-    return indices
-
-
-def _filter_model(atom_site, model_starts, model):
+def _filter_model(atom_site, model):
     """
     Reduce the ``atom_site`` category to the values for the given
     model.
+
+    Parameters
+    ----------
+    atom_site : CIFCategory or BinaryCIFCategory
+        ``atom_site`` category containing all models.
+    model : int
+        The model to be selected.
+
+    Returns
+    -------
+    atom_site : CIFCategory or BinaryCIFCategory
+        The ``atom_site`` category containing only the selected model.
     """
+    models = atom_site["pdbx_PDB_model_num"].as_array(np.int32)
+    _, model_starts = np.unique(models, return_index=True)
+    model_starts.sort()
     # Append exclusive stop
     model_starts = np.append(model_starts, [atom_site.row_count])
     # Indexing starts at 0, but model number starts at 1
@@ -703,7 +809,13 @@ def _get_box(block):
     return vectors_from_unitcell(len_a, len_b, len_c, alpha, beta, gamma)
 
 
-def set_structure(pdbx_file, array, data_block=None, include_bonds=False):
+def set_structure(
+    pdbx_file,
+    array,
+    data_block=None,
+    include_bonds=False,
+    extra_fields=[],
+):
     """
     Set the ``atom_site`` category with atom information from an
     :class:`AtomArray` or :class:`AtomArrayStack`.
@@ -737,6 +849,10 @@ def set_structure(pdbx_file, array, data_block=None, include_bonds=False):
         category.
         Inter-residue bonds will be written into the ``struct_conn``
         independent of this parameter.
+    extra_fields : list of str, optional
+        List of additional fields from the ``atom_site`` category
+        that should be written into the file.
+        Default is an empty list.
 
     Notes
     -----
@@ -752,7 +868,6 @@ def set_structure(pdbx_file, array, data_block=None, include_bonds=False):
     >>> file = CIFFile()
     >>> set_structure(file, atom_array)
     >>> file.write(os.path.join(path_to_directory, "structure.cif"))
-
     """
     _check_non_empty(array)
 
@@ -773,7 +888,11 @@ def set_structure(pdbx_file, array, data_block=None, include_bonds=False):
     )
     atom_site["label_comp_id"] = np.copy(array.res_name)
     atom_site["label_asym_id"] = np.copy(array.chain_id)
-    atom_site["label_entity_id"] = _determine_entity_id(array.chain_id)
+    atom_site["label_entity_id"] = (
+        np.copy(array.label_entity_id)
+        if "label_entity_id" in array.get_annotation_categories()
+        else _determine_entity_id(array.chain_id)
+    )
     atom_site["label_seq_id"] = np.copy(array.res_id)
     atom_site["pdbx_PDB_ins_code"] = Column(
         np.copy(array.ins_code),
@@ -797,6 +916,32 @@ def set_structure(pdbx_file, array, data_block=None, include_bonds=False):
             np.where(array.charge == 0, MaskValue.MISSING, MaskValue.PRESENT),
         )
 
+    # Handle all remaining custom fields
+    if len(extra_fields) > 0:
+        # ... check to avoid clashes with standard annotations
+        _standard_annotations = [
+            "hetero",
+            "element",
+            "atom_name",
+            "res_name",
+            "chain_id",
+            "res_id",
+            "ins_code",
+            "atom_id",
+            "b_factor",
+            "occupancy",
+            "charge",
+        ]
+        _reserved_annotation_names = list(atom_site.keys()) + _standard_annotations
+
+        for annot in extra_fields:
+            if annot in _reserved_annotation_names:
+                raise ValueError(
+                    f"Annotation name '{annot}' is reserved and cannot be written to as extra field. "
+                    "Please choose another name."
+                )
+            atom_site[annot] = np.copy(array.get_annotation(annot))
+
     if array.bonds is not None:
         struct_conn = _set_inter_residue_bonds(array, atom_site)
         if struct_conn is not None:
@@ -1021,13 +1166,21 @@ def _set_inter_residue_bonds(array, atom_site):
     if len(bond_array) == 0:
         return None
 
+    # Filter out 'standard' links, i.e. backbone bonds between adjacent canonical
+    # nucleotide/amino acid residues
+    bond_array = bond_array[~_filter_canonical_links(array, bond_array)]
+    if len(bond_array) == 0:
+        return None
+
     struct_conn = Category()
     struct_conn["id"] = np.arange(1, len(bond_array) + 1)
-    struct_conn["conn_type_id"] = np.full(len(bond_array), "covale")
+    struct_conn["conn_type_id"] = [
+        PDBX_BOND_TYPE_TO_TYPE_ID[btype] for btype in bond_array[:, 2]
+    ]
     struct_conn["pdbx_value_order"] = Column(
         np.array([PDBX_BOND_TYPE_TO_ORDER[btype] for btype in bond_array[:, 2]]),
         np.where(
-            bond_array[:, 2] == BondType.ANY,
+            np.isin(bond_array[:, 2], (BondType.ANY, BondType.COORDINATION)),
             MaskValue.MISSING,
             MaskValue.PRESENT,
         ),
@@ -1063,7 +1216,34 @@ def _filter_bonds(array, connection):
         raise ValueError("Invalid 'connection' option")
 
 
-def get_component(pdbx_file, data_block=None, use_ideal_coord=True, res_name=None):
+def _filter_canonical_links(array, bond_array):
+    """
+    Filter out peptide bonds between adjacent canonical amino acid residues.
+    """
+    # Get the residue index for each bonded atom
+    residue_indices = get_residue_positions(array, bond_array[:, :2].flatten()).reshape(
+        -1, 2
+    )
+
+    return (
+        # Must be canonical residues
+        np.isin(array.res_name[bond_array[:, 0]], CANONICAL_RESIDUE_LIST) &
+        np.isin(array.res_name[bond_array[:, 1]], CANONICAL_RESIDUE_LIST) &
+        # Must be backbone bond
+        np.isin(array.atom_name[bond_array[:, 0]], ("C", "O3'")) &
+        np.isin(array.atom_name[bond_array[:, 1]], ("N", "P")) &
+        # Must connect adjacent residues
+        residue_indices[:, 1] - residue_indices[:, 0] == 1
+    )  # fmt: skip
+
+
+def get_component(
+    pdbx_file,
+    data_block=None,
+    use_ideal_coord=True,
+    res_name=None,
+    allow_missing_coord=False,
+):
     """
     Create an :class:`AtomArray` for a chemical component from the
     ``chem_comp_atom`` and, if available, the ``chem_comp_bond``
@@ -1091,6 +1271,11 @@ def get_component(pdbx_file, data_block=None, use_ideal_coord=True, res_name=Non
         In this case, the component with the given residue name is
         read.
         By default, all rows would be read in this case.
+    allow_missing_coord : bool, optional
+        Whether to allow missing coordinate values in components.
+        If ``True``, these will be represented as ``nan`` values.
+        If ``False``, a ``ValueError`` is raised when missing coordinates
+        are encountered.
 
     Returns
     -------
@@ -1161,17 +1346,29 @@ def get_component(pdbx_file, data_block=None, use_ideal_coord=True, res_name=Non
         # Swap with the fallback option
         coord_fields, alt_coord_fields = alt_coord_fields, coord_fields
     try:
-        for i, field in enumerate(coord_fields):
-            array.coord[:, i] = atom_category[field].as_array(np.float32)
-    except KeyError as err:
-        key = err.args[0]
-        warnings.warn(
-            f"Attribute '{key}' not found within 'chem_comp_atom' category. "
-            f"The fallback coordinates will be used instead",
-            UserWarning,
+        array.coord = _parse_component_coordinates(
+            [atom_category[field] for field in coord_fields]
+        )
+    except Exception as err:
+        if isinstance(err, KeyError):
+            key = err.args[0]
+            warnings.warn(
+                f"Attribute '{key}' not found within 'chem_comp_atom' category. "
+                f"The fallback coordinates will be used instead",
+                UserWarning,
+            )
+        elif isinstance(err, ValueError):
+            warnings.warn(
+                "The coordinates are missing for some atoms. "
+                "The fallback coordinates will be used instead",
+                UserWarning,
+            )
+        else:
+            raise
+        array.coord = _parse_component_coordinates(
+            [atom_category[field] for field in alt_coord_fields],
+            allow_missing=allow_missing_coord,
         )
-        for i, field in enumerate(alt_coord_fields):
-            array.coord[:, i] = atom_category[field].as_array(np.float32)
 
     try:
         bond_category = block["chem_comp_bond"]
@@ -1181,7 +1378,7 @@ def get_component(pdbx_file, data_block=None, use_ideal_coord=True, res_name=Non
             )
     except KeyError:
         warnings.warn(
-            "Category 'chem_comp_bond' not found. " "No bonds will be parsed",
+            "Category 'chem_comp_bond' not found. No bonds will be parsed",
             UserWarning,
         )
     else:
@@ -1201,6 +1398,23 @@ def get_component(pdbx_file, data_block=None, use_ideal_coord=True, res_name=Non
     return array
 
 
+def _parse_component_coordinates(coord_columns, allow_missing=False):
+    coord = np.zeros((len(coord_columns[0]), 3), dtype=np.float32)
+    for i, column in enumerate(coord_columns):
+        if column.mask is not None and column.mask.array.any():
+            if allow_missing:
+                warnings.warn(
+                    "Missing coordinates for some atoms. Those will be set to nan",
+                    UserWarning,
+                )
+            else:
+                raise ValueError(
+                    "Missing coordinates for some atoms",
+                )
+        coord[:, i] = column.as_array(np.float32, masked_value=np.nan)
+    return coord
+
+
 def set_component(pdbx_file, array, data_block=None):
     """
     Set the ``chem_comp_atom`` and, if bonds are available,
@@ -1305,6 +1519,7 @@ def list_assemblies(pdbx_file, data_block=None):
 
     Examples
     --------
+
     >>> import os.path
     >>> file = CIFFile.read(os.path.join(path_to_structures, "1f2n.cif"))
     >>> assembly_ids = list_assemblies(file)
@@ -1417,7 +1632,10 @@ def get_assembly(
     Returns
     -------
     assembly : AtomArray or AtomArrayStack
-        The assembly. The return type depends on the `model` parameter.
+        The assembly.
+        The return type depends on the `model` parameter.
+        Contains the `sym_id` annotation, which enumerates the copies of the asymmetric
+        unit in the assembly.
 
     Examples
     --------
@@ -1506,7 +1724,6 @@ def _apply_transformations(structure, transformation_dict, operations):
     """
     # Additional first dimesion for 'structure.repeat()'
     assembly_coord = np.zeros((len(operations),) + structure.coord.shape)
-
     # Apply corresponding transformation for each copy in the assembly
     for i, operation in enumerate(operations):
         coord = structure.coord
@@ -1520,7 +1737,11 @@ def _apply_transformations(structure, transformation_dict, operations):
             coord += translation_vector
         assembly_coord[i] = coord
 
-    return repeat(structure, assembly_coord)
+    assembly = repeat(structure, assembly_coord)
+    assembly.set_annotation(
+        "sym_id", np.repeat(np.arange(len(operations)), structure.array_length())
+    )
+    return assembly
 
 
 def _get_transformations(struct_oper):
@@ -1596,4 +1817,118 @@ def _convert_string_to_sequence(string, stype):
     elif stype in _other_type_list:
         return None
     else:
-        raise InvalidFileError("mmCIF _entity_poly.type unsupported" " type: " + stype)
+        raise InvalidFileError("mmCIF _entity_poly.type unsupported type: " + stype)
+
+
+def get_sse(pdbx_file, data_block=None, match_model=None):
+    """
+    Get the secondary structure from a PDBx file.
+
+    Parameters
+    ----------
+    pdbx_file : CIFFile or CIFBlock or BinaryCIFFile or BinaryCIFBlock
+        The file object.
+        The following categories are required:
+
+        - ``entity_poly``
+        - ``struct_conf`` (if alpha-helices are present)
+        - ``struct_sheet_range`` (if beta-strands are present)
+        - ``atom_site`` (if `match_model` is set)
+
+    data_block : str, optional
+        The name of the data block.
+        Default is the first (and most times only) data block of the
+        file.
+        If the data block object is passed directly to `pdbx_file`,
+        this parameter is ignored.
+    match_model : None, optional
+        If a model number is given, only secondary structure elements for residues are
+        kept, that are resolved in the given model.
+        This means secondary structure elements for residues that would not appear
+        in a corresponding :class:`AtomArray` from :func:`get_structure()` are removed.
+        By default, all residues in the sequence are kept.
+
+    Returns
+    -------
+    sse_dict : dict of str -> ndarray, dtype=str
+        The dictionary maps the chain ID (derived from ``auth_asym_id``) to the
+        secondary structure of the respective chain.
+
+        - ``"a"``: alpha-helix
+        - ``"b"``: beta-strand
+        - ``"c"``: coil or not an amino acid
+
+        Each secondary structure element corresponds to the ``label_seq_id`` of the
+        ``atom_site`` category.
+        This means that the 0-th position of the array corresponds to the residue
+        in ``atom_site`` with ``label_seq_id`` ``1``.
+
+    Examples
+    --------
+
+    >>> import os.path
+    >>> file = CIFFile.read(os.path.join(path_to_structures, "1aki.cif"))
+    >>> sse = get_sse(file, match_model=1)
+    >>> print(sse)
+    {'A': array(['c', 'c', 'c', 'c', 'a', 'a', 'a', 'a', 'a', 'a', 'a', 'a', 'a',
+                 'a', 'c', 'c', 'c', 'c', 'c', 'a', 'a', 'a', 'c', 'c', 'a', 'a',
+                 'a', 'a', 'a', 'a', 'a', 'a', 'a', 'a', 'a', 'a', 'c', 'c', 'c',
+                 'c', 'c', 'c', 'b', 'b', 'b', 'c', 'c', 'c', 'c', 'c', 'b', 'b',
+                 'b', 'c', 'c', 'c', 'c', 'c', 'c', 'c', 'c', 'c', 'c', 'c', 'c',
+                 'c', 'c', 'c', 'c', 'c', 'c', 'c', 'c', 'c', 'c', 'c', 'c', 'c',
+                 'c', 'a', 'a', 'a', 'a', 'a', 'c', 'c', 'c', 'c', 'a', 'a', 'a',
+                 'a', 'a', 'a', 'a', 'a', 'a', 'a', 'a', 'a', 'a', 'c', 'c', 'a',
+                 'a', 'a', 'a', 'c', 'a', 'a', 'a', 'a', 'a', 'a', 'c', 'c', 'c',
+                 'c', 'c', 'a', 'a', 'a', 'a', 'c', 'c', 'c', 'c', 'c', 'c'],
+                 dtype='<U1')}
+
+    If only secondary structure elements for resolved residues are requested, the length
+    of the returned array matches the number of peptide residues in the structure.
+
+    >>> file = CIFFile.read(os.path.join(path_to_structures, "3o5r.cif"))
+    >>> print(len(get_sse(file, match_model=1)["A"]))
+    128
+    >>> atoms = get_structure(file, model=1)
+    >>> atoms = atoms[filter_amino_acids(atoms) & (atoms.chain_id == "A")]
+    >>> print(get_residue_count(atoms))
+    128
+    """
+    block = _get_block(pdbx_file, data_block)
+
+    # Init all chains with "c" for coil
+    sse_dict = {
+        chain_id: np.repeat("c", len(sequence))
+        for chain_id, sequence in get_sequence(block).items()
+    }
+
+    # Populate SSE arrays with helices and strands
+    for sse_symbol, category_name in [
+        ("a", "struct_conf"),
+        ("b", "struct_sheet_range"),
+    ]:
+        if category_name in block:
+            category = block[category_name]
+            chains = category["beg_auth_asym_id"].as_array(str)
+            start_positions = category["beg_label_seq_id"].as_array(int)
+            end_positions = category["end_label_seq_id"].as_array(int)
+
+            # set alpha helix positions
+            for chain, start, end in zip(chains, start_positions, end_positions):
+                # Translate the 1-based positions from PDBx into 0-based array indices
+                sse_dict[chain][start - 1 : end] = sse_symbol
+
+    if match_model is not None:
+        model_atom_site = _filter_model(block["atom_site"], match_model)
+        chain_ids = model_atom_site["auth_asym_id"].as_array(str)
+        res_ids = model_atom_site["label_seq_id"].as_array(int, masked_value=-1)
+        # Filter out masked residues, i.e. residues not part of a chain
+        mask = res_ids != -1
+        chain_ids = chain_ids[mask]
+        res_ids = res_ids[mask]
+        for chain_id, sse in sse_dict.items():
+            res_ids_in_chain = res_ids[chain_ids == chain_id]
+            # Transform from 1-based residue ID to 0-based index
+            indices = np.unique(res_ids_in_chain) - 1
+            sse_dict[chain_id] = sse[indices]
+
+    return sse_dict