File size: 8,557 Bytes
f2ad668 | 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 | from itertools import groupby
from typing import Any
import numpy as np
import torch
from .esmfold2_constants import ELEMENT_NUMBER_TO_SYMBOL, MOL_TYPE_NONPOLYMER
from .esmfold2_molecular_complex import (
MolecularComplex,
MolecularComplexMetadata,
)
def get_element_symbol(atomic_num: int) -> str:
return ELEMENT_NUMBER_TO_SYMBOL.get(atomic_num, "X")
def build_molecular_complex_from_features(
coords: torch.Tensor,
plddt: torch.Tensor,
atom_mask: torch.Tensor,
ref_element: torch.Tensor,
ref_atom_name_chars: torch.Tensor,
chain_infos: list,
complex_id: str,
) -> MolecularComplex:
"""Construct a MolecularComplex from feature-dict tensors and chain metadata.
Non-polymer chains (ligands) collapse all per-atom tokens into a single
residue token whose pLDDT is the per-token average and whose hetero flag
is True.
"""
mask_np = atom_mask.bool().cpu().numpy()
coords_np = coords.float().cpu().numpy()
name_chars_np = ref_atom_name_chars.cpu().numpy()
elements_np = ref_element.cpu().numpy()
plddt_np = plddt.float().cpu().numpy()
sequence_tokens: list[str] = []
chain_ids_per_token: list[int] = []
token_to_atoms: list[list[int]] = []
confidence: list[float] = []
flat_positions: list[list[float]] = []
flat_elements: list[str] = []
flat_names: list[str] = []
flat_hetero: list[bool] = []
chain_lookup: dict[int, str] = {}
entity_info: dict[int, str] = {}
out_atom_cursor = 0
for ci in chain_infos:
chain_lookup[ci.asym_id] = ci.chain_id
is_nonpolymer = ci.mol_type == MOL_TYPE_NONPOLYMER
entity_info[ci.entity_id] = "non-polymer" if is_nonpolymer else "polymer"
if is_nonpolymer:
residue_name = ci.tokens[0].residue_name if ci.tokens else "LIG"
sequence_tokens.append(residue_name)
chain_ids_per_token.append(ci.asym_id)
avg_plddt = (
float(np.mean([plddt_np[ti.token_index] for ti in ci.tokens]))
if ci.tokens
else 0.0
)
confidence.append(avg_plddt)
token_atom_start = out_atom_cursor
for ti in ci.tokens:
for atom_idx in range(ti.atom_start, ti.atom_start + ti.atom_count):
if not mask_np[atom_idx]:
continue
flat_positions.append(coords_np[atom_idx].tolist())
flat_elements.append(get_element_symbol(int(elements_np[atom_idx])))
chars = name_chars_np[atom_idx]
name = "".join(
chr(int(c) + 32) for c in chars if int(c) != 0
).strip()
flat_names.append(name)
flat_hetero.append(True)
out_atom_cursor += 1
token_to_atoms.append([token_atom_start, out_atom_cursor])
continue
# Atom-tokenized modified residues (HYP, MSE, ...) span multiple
# tokens per residue; collapse them back to one mmCIF residue.
for _residue_index, ti_iter in groupby(
ci.tokens, key=lambda t: t.residue_index
):
ti_group = list(ti_iter)
sequence_tokens.append(ti_group[0].residue_name)
chain_ids_per_token.append(ci.asym_id)
confidence.append(
float(np.mean([plddt_np[ti.token_index] for ti in ti_group]))
)
token_atom_start = out_atom_cursor
for ti in ti_group:
for atom_idx in range(ti.atom_start, ti.atom_start + ti.atom_count):
if not mask_np[atom_idx]:
continue
flat_positions.append(coords_np[atom_idx].tolist())
flat_elements.append(get_element_symbol(int(elements_np[atom_idx])))
chars = name_chars_np[atom_idx]
name = "".join(
chr(int(c) + 32) for c in chars if int(c) != 0
).strip()
flat_names.append(name)
flat_hetero.append(False)
out_atom_cursor += 1
token_to_atoms.append([token_atom_start, out_atom_cursor])
return MolecularComplex(
id=complex_id,
sequence=sequence_tokens,
atom_positions=np.array(flat_positions, dtype=np.float32).reshape(-1, 3),
atom_elements=np.array(flat_elements, dtype=object),
token_to_atoms=np.array(token_to_atoms, dtype=np.int32).reshape(-1, 2),
chain_id=np.array(chain_ids_per_token, dtype=np.int64),
plddt=np.array(confidence, dtype=np.float32),
atom_names=np.array(flat_names, dtype=object),
atom_hetero=np.array(flat_hetero, dtype=bool),
metadata=MolecularComplexMetadata(
entity_lookup=entity_info,
chain_lookup=chain_lookup,
assembly_composition=None,
),
)
def build_molecular_complex(
structure: Any, coords: torch.Tensor, plddt: torch.Tensor, complex_id: str
) -> MolecularComplex:
"""Directly constructs a MolecularComplex from model outputs without intermediate files.
Args:
structure: Object with .chains, .residues, .atoms numpy structured arrays.
coords: [N_atoms, 3] predicted atom coordinates.
plddt: [N_residues] per-residue confidence scores.
complex_id: Identifier string for the resulting complex.
"""
flat_positions = []
flat_elements = []
flat_names = []
flat_hetero = []
sequence_tokens = []
token_to_atoms = []
chain_ids_per_token = []
confidence_scores = []
chain_lookup = {}
entity_info = {}
global_atom_cursor = 0
global_res_cursor = 0
atom_array_idx = 0
for chain in structure.chains:
chain_idx_numeric = chain["asym_id"]
chain_name_str = str(chain["name"])
mol_type = chain["mol_type"]
chain_lookup[chain_idx_numeric] = chain_name_str
entity_info[chain["entity_id"]] = (
"polymer" if mol_type != MOL_TYPE_NONPOLYMER else "non-polymer"
)
res_start = chain["res_idx"]
res_end = chain["res_idx"] + chain["res_num"]
residues = structure.residues[res_start:res_end]
for residue in residues:
res_name = str(residue["name"])
sequence_tokens.append(res_name)
chain_ids_per_token.append(chain_idx_numeric)
score = plddt[global_res_cursor].item()
confidence_scores.append(score)
token_start_idx = atom_array_idx
atom_start = residue["atom_idx"]
atom_end = residue["atom_idx"] + residue["atom_num"]
atoms = structure.atoms[atom_start:atom_end]
for atom in atoms:
if not atom["is_present"]:
continue
pos = coords[global_atom_cursor].tolist()
flat_positions.append(pos)
elem = get_element_symbol(atom["element"].item())
flat_elements.append(elem)
raw_name = atom["name"]
if hasattr(raw_name, "tolist"):
raw_name = raw_name.tolist()
name_str = "".join([chr(c + 32) for c in raw_name if c != 0])
flat_names.append(name_str)
flat_hetero.append(mol_type == MOL_TYPE_NONPOLYMER)
global_atom_cursor += 1
atom_array_idx += 1
token_to_atoms.append([token_start_idx, atom_array_idx])
global_res_cursor += 1
return MolecularComplex(
id=complex_id,
sequence=sequence_tokens,
atom_positions=np.array(flat_positions, dtype=np.float32),
atom_elements=np.array(flat_elements, dtype=object),
token_to_atoms=np.array(token_to_atoms, dtype=np.int32),
chain_id=np.array(chain_ids_per_token, dtype=np.int64),
plddt=np.array(confidence_scores, dtype=np.float32),
atom_names=np.array(flat_names, dtype=object),
atom_hetero=np.array(flat_hetero, dtype=bool),
metadata=MolecularComplexMetadata(
entity_lookup=entity_info,
chain_lookup=chain_lookup,
assembly_composition=None,
),
)
|