File size: 18,681 Bytes
13e95c8 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 | """
Data interface definitions for ODesign dataset and feature processing.
This module defines standardized data structures for features and labels:
- OFeatureData: Complete feature set for model input
- OLabelData: Ground truth labels for training
These interfaces ensure type consistency and provide convenient property
accessors for data dimensions.
"""
from ._base import *
from typing import List
from attr import define
import attr
import numpy as np
from src.utils.license_register import register_license
@define(kw_only=True)
@register_license('odesign2025')
@auto_type_convert
class OFeatureData(DictAccessMixin):
"""
Complete feature data structure for ODesign model input.
This class contains all features extracted from protein/RNA/DNA/ligand structures,
including token-level features, atomic features, MSA features, template features,
and various masks for training and evaluation. Features are organized by category.
Attributes:
# Token-level features
token_index (IndexType): Token index within the full sequence.
Shape: [N_token]
residue_index (IndexType): Residue index in original sequence (1-indexed).
Shape: [N_token]
asym_id (IndexType): Asymmetric unit ID (chain identifier).
Shape: [N_token]
entity_id (IndexType): Entity ID (molecule type identifier).
Same entity ID = same sequence.
Shape: [N_token]
sym_id (IndexType): Symmetry operation ID.
Shape: [N_token]
restype (OneHotType): One-hot encoded residue/nucleotide/ligand type.
Shape: [N_token, n_residue_types]
# Bond features
token_bonds (FeatureType): Bond features between token pairs.
Encodes bond types and connectivity.
Shape: [N_token, N_token, n_bond_features]
token_pair_gen_mask (MaskType): Mask for token pairs to generate/evaluate.
Shape: [N_token, N_token]
token_bond_gen_mask (MaskType): Mask for bond type generation/evaluation.
Shape: [N_token, N_token]
# Reference atomic features
ref_pos (FeatureType): Reference atomic positions from structure.
Shape: [N_atom, 3]
ref_mask (MaskType): Mask for valid reference atoms (exists in structure).
Shape: [N_atom]
ref_element (FeatureType): Atomic element encoding.
Shape: [N_atom, n_elements] or [N_atom]
ref_charge (FeatureType): Atomic partial charge.
Shape: [N_atom]
ref_atom_name_chars (OneHotType): One-hot encoded atom name characters.
Shape: [N_atom, 4, n_char_types]
ref_space_uid (IndexType): Spatial unit ID for grouping atoms in same frame.
Used for residue-level symmetric atom permutation.
Shape: [N_atom]
has_frame (MaskType): Mask indicating atoms with valid coordinate frames.
Shape: [N_atom]
frame_atom_index (IndexType): Indices of atoms defining coordinate frames.
For proteins: [N, CA, C] atoms; for nucleotides: [C1', C3', C4']
Shape: [N_frame, 3]
# Identity and mapping features
atom_to_token_idx (IndexType): Mapping from atom index to token index.
Shape: [N_atom]
atom_to_tokatom_idx (IndexType): Mapping from atom to token-atom index.
Index within atoms of the same token.
Shape: [N_atom]
is_protein (MaskType): Binary mask indicating protein atoms.
Shape: [N_atom]
is_ligand (MaskType): Binary mask indicating ligand atoms.
Shape: [N_atom]
is_dna (MaskType): Binary mask indicating DNA atoms.
Shape: [N_atom]
is_rna (MaskType): Binary mask indicating RNA atoms.
Shape: [N_atom]
# Resolution
resolution (FeatureType): Structure resolution in Angstroms.
Used for filtering confidence loss. Shape: scalar or [1]
# Chain permutation features
mol_id (IndexType): Molecule ID for each atom.
Identifies individual molecules (multiple copies of same entity).
Shape: [N_atom]
mol_atom_index (IndexType): Atom index within molecule.
Shape: [N_atom]
entity_mol_id (IndexType): Entity molecule ID for each atom.
Shape: [N_atom]
masked_asym_ids (List[int], optional): List of asymmetric unit IDs
to consider for chain permutation. None means all chains.
# Atom permutation features
atom_perm_list (list[list[int]]): List of symmetric atom groups.
Each group contains atom indices that can be permuted
(e.g., symmetric atoms in aromatic rings, carboxyl groups).
# Various masks for training and evaluation
pae_rep_atom_mask (MaskType): Representative atom mask for PAE computation.
Shape: [N_atom]
modified_res_mask (MaskType): Mask for modified residues.
Shape: [N_token]
is_condition_atom (MaskType): Mask for conditioned (fixed) atoms.
Atoms not generated by diffusion.
Shape: [N_atom]
distogram_rep_atom_mask (MaskType): Representative atom mask for distogram.
Typically Cβ/Cα for proteins, C4'/C2' for nucleotides.
Shape: [N_atom]
plddt_m_rep_atom_mask (MaskType): Representative atom mask for pLDDT-m.
Shape: [N_atom]
bond_mask (MaskType): Mask for valid bonds.
Shape: [N_atom, N_atom] or [N_token, N_token]
# Constraint features
constraint_feature (FeatureType, optional): Constraint features for token pairs.
Currently supports distogram constraint, pocket constraint, etc.
Shape: [N_token, N_token, n_constraint_features]
# MSA features (optional)
msa (IndexType, optional): Multiple sequence alignment.
Token indices from MSA sequences.
Shape: [N_msa, N_token]
has_deletion (MaskType, optional): Mask indicating deletions in MSA.
Shape: [N_msa, N_token]
deletion_value (FeatureType, optional): Deletion values in MSA.
Shape: [N_msa, N_token]
profile (FeatureType, optional): Sequence profile from MSA.
Position-specific scoring matrix.
Shape: [N_token, 32]
deletion_mean (FeatureType, optional): Mean deletion probability.
Shape: [N_token]
msa_token_mask (MaskType, optional): Mask for valid MSA tokens.
Shape: [N_msa, N_token]
prot_pair_num_alignments (torch.Tensor, optional): Number of paired
protein alignments.
prot_unpair_num_alignments (torch.Tensor, optional): Number of unpaired
protein alignments.
rna_pair_num_alignments (torch.Tensor, optional): Number of paired
RNA alignments.
rna_unpair_num_alignments (torch.Tensor, optional): Number of unpaired
RNA alignments.
# Template features (under development)
template_restype (IndexType, optional): Residue types in template structures.
Shape: [N_template, N_token]
template_all_atom_mask (MaskType, optional): Atom masks in templates.
Shape: [N_template, N_token, n_atom_per_token]
template_all_atom_positions (FeatureType, optional): Atom positions in templates.
Shape: [N_template, N_token, n_atom_per_token, 3]
# Hotspot features
is_hotspot_residue (FeatureType, optional): Binary mask for hotspot residues.
Residues of particular interest (e.g., binding site, catalytic residues).
Shape: [N_token]
# Cyclic structure features
is_cyclic_token (FeatureType, optional): Binary mask for cyclic peptide tokens.
Shape: [N_token]
Properties:
num_asym: Number of asymmetric units (chains).
num_token: Number of tokens.
num_atom: Number of atoms.
num_msa: Number of MSA sequences.
default_num_msa: Default number of MSA sequences for padding (1).
default_num_templ: Default number of templates for padding (4).
default_num_pocket: Default number of pocket residues (30).
"""
# Token-level features
token_index: IndexType
residue_index: IndexType
asym_id: IndexType
entity_id: IndexType
sym_id: IndexType
restype: OneHotType
# Bond features
token_bonds: FeatureType
token_pair_gen_mask: MaskType
token_bond_gen_mask: MaskType
# Reference atomic features
ref_pos: FeatureType
ref_mask: MaskType
ref_element: FeatureType
ref_charge: FeatureType
ref_atom_name_chars: OneHotType
ref_space_uid: IndexType
has_frame: MaskType
frame_atom_index: IndexType
# Identity and mapping features
atom_to_token_idx: IndexType
atom_to_tokatom_idx: IndexType
is_protein: MaskType
is_ligand: MaskType
is_dna: MaskType
is_rna: MaskType
# Resolution
resolution: FeatureType
# Chain permutation features
mol_id: IndexType
mol_atom_index: IndexType
entity_mol_id: IndexType
masked_asym_ids: List[int] | None = None
# Atom permutation features
atom_perm_list: list[list[int]]
# Mask features for training and evaluation
pae_rep_atom_mask: MaskType
modified_res_mask: MaskType
is_condition_atom: MaskType
distogram_rep_atom_mask: MaskType
plddt_m_rep_atom_mask: MaskType
bond_mask: MaskType
# Constraint features
constraint_feature: FeatureType | None = None
# MSA features (optional)
msa: IndexType | None = None
has_deletion: MaskType | None = None
deletion_value: FeatureType | None = None
profile: FeatureType | None = None
deletion_mean: FeatureType | None = None
msa_token_mask: MaskType | None = None
prot_pair_num_alignments: torch.Tensor | None = None
prot_unpair_num_alignments: torch.Tensor | None = None
rna_pair_num_alignments: torch.Tensor | None = None
rna_unpair_num_alignments: torch.Tensor | None = None
# Template features (under development)
template_restype: IndexType | None = None
template_all_atom_mask: MaskType | None = None
template_all_atom_positions: FeatureType | None = None
# Hotspot features
is_hotspot_residue: FeatureType | None = None
# Cyclic structure features
is_cyclic_token: FeatureType | None = None
@classmethod
def from_feature_dict(cls, feature_dict: dict) -> "OFeatureData":
"""
Create OFeatureData instance from a feature dictionary.
This factory method filters the input dictionary to only include
fields defined in OFeatureData and validates that all required
fields are present.
Args:
feature_dict (dict): Dictionary containing feature data.
Keys should match OFeatureData field names.
Returns:
OFeatureData: Validated feature data instance.
Raises:
AssertionError: If required fields are missing.
"""
all_fields = {field.name for field in attr.fields(cls)}
filtered_dict = {k: v for k, v in feature_dict.items() if k in all_fields}
def _check_required_fields(filtered_dict):
required_fields = {
field.name for field in attr.fields(cls)
if field.default == attr.NOTHING
}
missing_fields = required_fields - set(filtered_dict.keys())
assert not missing_fields, (
f"Missing required fields in {cls.__name__}: {missing_fields}"
)
_check_required_fields(filtered_dict)
return cls(**filtered_dict)
@property
def num_asym(self) -> int:
"""
Get the number of asymmetric units (chains) in the structure.
Returns:
int: Number of unique chains.
"""
return len(torch.unique(self.asym_id))
@property
def num_token(self) -> int:
"""
Get the number of tokens in the sequence.
Returns:
int: Number of tokens (residues/nucleotides/ligand atoms).
"""
return self.token_index.shape[0]
@property
def num_atom(self) -> int:
"""
Get the number of atoms in the structure.
Returns:
int: Total number of atoms.
"""
return self.atom_to_token_idx.shape[0]
@property
def num_msa(self) -> int:
"""
Get the number of MSA sequences.
Returns:
int: Number of sequences in multiple sequence alignment.
"""
return self.msa.shape[0]
@property
def default_num_msa(self) -> int:
"""
Get the default number of MSA sequences for padding.
Returns:
int: Default value (1) for structures without MSA.
"""
return 1
@property
def default_num_templ(self) -> int:
"""
Get the default number of template structures for padding.
Returns:
int: Default value (4) for template feature dimensions.
"""
return 4
@property
def default_num_pocket(self) -> int:
"""
Get the default number of pocket residues.
Returns:
int: Default value (30) for pocket size in ligand binding.
"""
return 30
@define(kw_only=True)
@register_license('odesign2025')
@auto_type_convert
class OLabelData(DictAccessMixin):
"""
Ground truth label data structure for training and evaluation.
This class contains all ground truth labels including atomic coordinates,
bond types, and metadata needed for loss computation and evaluation metrics.
Attributes:
coordinate (FeatureType): Ground truth atomic coordinates.
Reference positions from experimental structure.
Shape: [N_atom, 3]
coordinate_mask (MaskType): Mask indicating which atoms have valid coordinates.
Used to filter missing or disordered atoms.
Shape: [N_atom]
token_bond_type_label (FeatureType, optional): Ground truth bond type labels.
Encodes bond types for token pairs (no_bond, single, double, triple, aromatic).
Shape: [N_token, N_token] or [N_token, N_token, n_bond_types]
ligand_bond_mask (MaskType, optional): Mask for ligand bonds.
Indicates which bonds should be considered for ligand bond loss.
Shape: [N_atom, N_atom] or [N_token, N_token]
entity_mol_id (IndexType, optional): Entity molecule ID for each atom.
Used for chain permutation during training.
Shape: [N_atom]
mol_id (IndexType, optional): Molecule ID for each atom.
Identifies individual molecule instances.
Shape: [N_atom]
mol_atom_index (IndexType, optional): Atom index within molecule.
Maps atoms to their position within molecule for permutation.
Shape: [N_atom]
pae_rep_atom_mask (MaskType, optional): Representative atom mask for PAE.
Identifies atoms used as representatives for predicted aligned error.
Shape: [N_atom]
eval_type (np.ndarray, optional): Evaluation type identifier.
Specifies the type of evaluation (e.g., interface, binding, folding).
cluster_id (np.ndarray, optional): Cluster ID for evaluation.
Used to group similar structures in validation.
chain_1_mask (MaskType, optional): Mask for chain 1 in interface evaluation.
Used for protein-protein interface or ligand binding evaluation.
Shape: [N_atom]
chain_2_mask (MaskType, optional): Mask for chain 2 in interface evaluation.
Used for protein-protein interface or ligand binding evaluation.
Shape: [N_atom]
"""
coordinate: FeatureType
coordinate_mask: MaskType
token_bond_type_label: FeatureType | None = None
ligand_bond_mask: MaskType | None = None
entity_mol_id: IndexType | None = None
mol_id: IndexType | None = None
mol_atom_index: IndexType | None = None
pae_rep_atom_mask: MaskType | None = None
eval_type: np.ndarray | None = None
cluster_id: np.ndarray | None = None
chain_1_mask: MaskType | None = None
chain_2_mask: MaskType | None = None
@classmethod
def from_label_dict(cls, label_dict: dict) -> "OLabelData":
"""
Create OLabelData instance from a label dictionary.
This factory method filters the input dictionary to only include
fields defined in OLabelData and validates that all required
fields are present.
Args:
label_dict (dict): Dictionary containing label data.
Keys should match OLabelData field names.
Returns:
OLabelData: Validated label data instance.
Raises:
AssertionError: If required fields are missing.
"""
all_fields = {field.name for field in attr.fields(cls)}
filtered_dict = {k: v for k, v in label_dict.items() if k in all_fields}
def _check_required_fields(filtered_dict):
required_fields = {
field.name for field in attr.fields(cls)
if field.default == attr.NOTHING
}
missing_fields = required_fields - set(filtered_dict.keys())
assert not missing_fields, (
f"Missing required fields in {cls.__name__}: {missing_fields}"
)
_check_required_fields(filtered_dict)
return cls(**filtered_dict)
|