dualmapref 0.2.8__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- args_parse.py +593 -0
- candidate_corrector.py +161 -0
- candidate_scorer.py +359 -0
- conformend/__init__.py +36 -0
- conformend/refiner.py +25 -0
- dataset/__init__.py +1 -0
- dataset/drugdataset.py +644 -0
- dataset/manifold.py +237 -0
- dataset/sharded_drugdataset.py +124 -0
- dataset/smiles2graph.py +217 -0
- dualmapref/__init__.py +24 -0
- dualmapref/refiner.py +16 -0
- dualmapref-0.2.8.dist-info/LICENSE +21 -0
- dualmapref-0.2.8.dist-info/METADATA +167 -0
- dualmapref-0.2.8.dist-info/RECORD +186 -0
- dualmapref-0.2.8.dist-info/WHEEL +5 -0
- dualmapref-0.2.8.dist-info/entry_points.txt +10 -0
- dualmapref-0.2.8.dist-info/top_level.txt +29 -0
- exputils.py +289 -0
- infer.py +23 -0
- infer_base_conformers.py +5589 -0
- infer_mean_dispersion_torsion.py +1117 -0
- infer_torsion.py +752 -0
- infer_torsional_denoiser.py +1021 -0
- manimol/__init__.py +30 -0
- manimol/__main__.py +6 -0
- manimol/cli.py +36 -0
- manimol/full_cli.py +41 -0
- manimol/legacy_imports.py +43 -0
- manimol/manifest.py +56 -0
- manimol/postprocess.py +148 -0
- manimol/priors.py +60 -0
- manimol/protocols.py +119 -0
- manimol/refiner.py +645 -0
- manimol/runtime.py +194 -0
- manimol/selection.py +85 -0
- manimol/torsion.py +22 -0
- manimol/train_cli.py +14 -0
- mixture_flow/src/__init__.py +1 -0
- mixture_flow/src/_path.py +7 -0
- mixture_flow/src/infer28_mixture_torsion_flow.py +1093 -0
- mixture_flow/src/mixture_torsion_flow.py +291 -0
- mixture_flow/src/summarize_mixture_debug.py +80 -0
- mixture_flow/src/train28_mixture_torsion_flow.py +583 -0
- mixture_torsion_flow.py +7 -0
- models/ manifold_learning.py +86 -0
- models/__init__.py +8 -0
- models/dist.py +273 -0
- models/dist2coords.py +23 -0
- models/egnn.py +107 -0
- models/experimental_backbones.py +364 -0
- models/gnnconv.py +512 -0
- models/kernel_inversion.py +141 -0
- models/kernels.py +427 -0
- models/losses.py +154 -0
- models/mean_dispersion_bridge.py +119 -0
- models/mean_dispersion_torsion_context.py +155 -0
- models/model.py +73 -0
- models/model0.py +2375 -0
- models/vis.py +128 -0
- pgraph_controls.py +107 -0
- scripts/__init__.py +2 -0
- scripts/analyze_basin_alignment.py +63 -0
- scripts/analyze_hif2a_conformer_aggregation.py +267 -0
- scripts/analyze_hif2a_contact_pharmacophore.py +1111 -0
- scripts/analyze_hif2a_fixed_budget.py +202 -0
- scripts/analyze_hif2a_pharmacophore.py +405 -0
- scripts/analyze_hif2a_pharmacophore_match_metric.py +386 -0
- scripts/analyze_p_control_candidate_flow.py +154 -0
- scripts/analyze_scaffold_basin.py +83 -0
- scripts/audit_dispersion_proposals.py +208 -0
- scripts/audit_etflow_protocol.py +258 -0
- scripts/audit_external_overlap.py +150 -0
- scripts/audit_oldbest_fulltrainval_split.py +82 -0
- scripts/audit_pdbbind_contact_pharmacophores.py +146 -0
- scripts/audit_qm9_geomol_test_split.py +94 -0
- scripts/audit_stage1_stage2_cross.py +131 -0
- scripts/build_geom_drugs_lowref_diagnostic.py +84 -0
- scripts/build_qm9_backfill_manifest.py +99 -0
- scripts/check_etflow_runtime.py +56 -0
- scripts/compare_prior_sources.py +122 -0
- scripts/compose_binding_mode_figure.py +87 -0
- scripts/compose_hif2a_contact_pharmacophore_suppfig.py +255 -0
- scripts/compose_hif2a_functional_geometry_pipeline.py +266 -0
- scripts/compose_manimol_internal_architecture.py +291 -0
- scripts/diagnose_flow_basin_intervention.py +54 -0
- scripts/diagnose_hif2a_pharmacophore_budget.py +185 -0
- scripts/diagnose_route2_p_ranking.py +123 -0
- scripts/discover_binding_mode_inputs.py +30 -0
- scripts/evaluate_functional_prior_stage1.py +230 -0
- scripts/evaluate_geomanimol.py +74 -0
- scripts/evaluate_geomanimol_flow.py +23 -0
- scripts/evaluate_geomanimol_multiscaffold.py +46 -0
- scripts/evaluate_geomanimol_residual.py +45 -0
- scripts/evaluate_geomanimol_torsion.py +40 -0
- scripts/evaluate_hif2a_bioactive_enrichment.py +91 -0
- scripts/evaluate_hif2a_pms_suite.py +440 -0
- scripts/evaluate_pdbbind_contact_pms.py +155 -0
- scripts/evaluate_platinum_external.py +114 -0
- scripts/evaluate_platinum_pool.py +82 -0
- scripts/evaluate_route2_calibrated_p.py +104 -0
- scripts/evaluate_route2_library_components.py +596 -0
- scripts/evaluate_route2_p_basin_budget.py +288 -0
- scripts/evaluate_stage1_checkpoint_only.py +30 -0
- scripts/export_dispersion_bond_audit.py +136 -0
- scripts/export_top20_shards.py +207 -0
- scripts/external_validation_metrics.py +152 -0
- scripts/fit_topology_prior.py +117 -0
- scripts/frozen_dispersion_audit.py +281 -0
- scripts/geomanimol_debug.py +50 -0
- scripts/geomanimol_smoke.py +37 -0
- scripts/infer_etflow_matched.py +524 -0
- scripts/inspect_checkpoint.py +12 -0
- scripts/inspect_conformer_capacity.py +71 -0
- scripts/inspect_pool.py +12 -0
- scripts/make_figure3_selection_loss.py +474 -0
- scripts/make_figure3_selection_summary.py +275 -0
- scripts/make_geom_debug_subset.py +62 -0
- scripts/make_low_reference_test_subset.py +70 -0
- scripts/merge_stage1_into_stage2.py +82 -0
- scripts/place_hif2a_candidates_for_evaluation.py +66 -0
- scripts/plot_hif2a_low_budget_efficiency.py +100 -0
- scripts/plot_prior_validation.py +48 -0
- scripts/precompute_stage2_assignments.py +90 -0
- scripts/prepare_bace_activity.py +98 -0
- scripts/prepare_bace_etkdg_benchmark.py +90 -0
- scripts/prepare_figure5_4x4_cases.py +470 -0
- scripts/prepare_geom_drugs_geomol_top20.py +472 -0
- scripts/prepare_pdbbind_core_raw.py +90 -0
- scripts/prepare_platinum_raw.py +74 -0
- scripts/prepare_pubchem_etkdg.py +54 -0
- scripts/rebuild_hif2a_final_sdf_and_metrics.py +55 -0
- scripts/rebuild_hif2a_fixed_pool_sweep.py +175 -0
- scripts/rebuild_hif2a_strict.py +220 -0
- scripts/recompute_hif2a_bootstrap.py +63 -0
- scripts/reference_free_geometry.py +112 -0
- scripts/render_binding_modes.py +413 -0
- scripts/render_figure5_4x4.py +330 -0
- scripts/render_figure6_pharmacophore_pymol.py +162 -0
- scripts/render_hif2a_contact_pharmacophore_pymol.py +414 -0
- scripts/render_hif2a_figure11.py +397 -0
- scripts/run_bace_2d_baseline.py +73 -0
- scripts/run_drugs_validation_matrix.py +225 -0
- scripts/run_external_reference_free.py +53 -0
- scripts/run_manimol_finalk_sensitivity.py +106 -0
- scripts/run_manimol_oversampling_sensitivity.py +109 -0
- scripts/run_pdisp_matched_intervention.py +116 -0
- scripts/run_stage1_stage2_cross20.py +83 -0
- scripts/score_hif2a_candidates_with_pgraph.py +221 -0
- scripts/summarize_dispersion_intervention.py +88 -0
- scripts/test_mean_dispersion_bridge.py +50 -0
- scripts/train_dispersion_head_diagnostic.py +320 -0
- scripts/train_dispersion_head_streaming.py +519 -0
- scripts/train_geomanimol.py +100 -0
- scripts/train_geomanimol_flow.py +26 -0
- scripts/train_geomanimol_residual.py +70 -0
- scripts/train_geomanimol_torsion.py +50 -0
- scripts/train_quantile_representation_diagnostic.py +204 -0
- scripts/validate_mean_dispersion_prior.py +173 -0
- scripts/validate_packed_shard.py +41 -0
- scripts/validate_stage1_prior.py +221 -0
- scripts/validate_top20_shards.py +36 -0
- scripts/validation/__init__.py +1 -0
- scripts/validation/hif2a_downstream_baselines.py +105 -0
- scripts/validation/hif2a_pairwise_ranker.py +218 -0
- scripts/verify_binding_mode_figure.py +47 -0
- torsion_manifold.py +281 -0
- train.py +78 -0
- train_graph_encoder.py +206 -0
- train_local_torsion_corrector.py +511 -0
- train_mean_dispersion_torsion_prior.py +329 -0
- train_probability_basin.py +235 -0
- train_torsion_generator.py +1631 -0
- train_torsion_manifold_prior.py +258 -0
- train_torsional_denoiser.py +415 -0
- trainer.py +1038 -0
- utils/__init__.py +7 -0
- utils/checkpoint.py +40 -0
- utils/device.py +17 -0
- utils/helpers.py +98 -0
- utils/kabsch.py +29 -0
- utils/lookup_table.py +164 -0
- utils/metrics.py +50 -0
- utils/optuna.py +111 -0
- utils/save_mol.py +366 -0
- utils/util.py +36 -0
dataset/manifold.py
ADDED
|
@@ -0,0 +1,237 @@
|
|
|
1
|
+
import numpy as np
|
|
2
|
+
from typing import Optional, Dict, Any
|
|
3
|
+
from models import (
|
|
4
|
+
compute_AE_tanimoto_distance_np,
|
|
5
|
+
compute_augmented_graph_distance_np,
|
|
6
|
+
compute_embed3d_distance_np,
|
|
7
|
+
hop_matrix_from_mol,
|
|
8
|
+
)
|
|
9
|
+
|
|
10
|
+
|
|
11
|
+
def _extract_distance_matrix(dist_used, prefer_key: Optional[str] = "D_aug"):
|
|
12
|
+
"""从可能的返回类型(np.ndarray / dict / list-like)中提取方阵距离矩阵。
|
|
13
|
+
|
|
14
|
+
优先查找 prefer_key,然后查找常见键名,再回退到第一个能识别为方阵的 array-like 值。
|
|
15
|
+
如果无法提取,将抛出 ValueError。
|
|
16
|
+
"""
|
|
17
|
+
if isinstance(dist_used, dict):
|
|
18
|
+
# 优先取 prefer_key
|
|
19
|
+
if prefer_key is not None and prefer_key in dist_used:
|
|
20
|
+
return np.asarray(dist_used[prefer_key], dtype=float)
|
|
21
|
+
|
|
22
|
+
# 常见候选键
|
|
23
|
+
for key in ("D_aug", "D_topo", "D_edge", "D_feat", "dist", "matrix", "D"):
|
|
24
|
+
if key in dist_used:
|
|
25
|
+
return np.asarray(dist_used[key], dtype=float)
|
|
26
|
+
|
|
27
|
+
# 回退:第一个能解析成方阵的 value
|
|
28
|
+
for v in dist_used.values():
|
|
29
|
+
try:
|
|
30
|
+
arr = np.asarray(v, dtype=float)
|
|
31
|
+
except Exception:
|
|
32
|
+
continue
|
|
33
|
+
if arr.ndim == 2 and arr.shape[0] == arr.shape[1]:
|
|
34
|
+
return arr
|
|
35
|
+
|
|
36
|
+
raise ValueError(
|
|
37
|
+
"Cannot extract a square distance matrix from dict. Available keys: %s"
|
|
38
|
+
% (list(dist_used.keys()),)
|
|
39
|
+
)
|
|
40
|
+
else:
|
|
41
|
+
arr = np.asarray(dist_used, dtype=float)
|
|
42
|
+
if arr.ndim != 2 or arr.shape[0] != arr.shape[1]:
|
|
43
|
+
raise ValueError("dist must be a square (n x n) array, got shape: %s" % (arr.shape,))
|
|
44
|
+
return arr
|
|
45
|
+
|
|
46
|
+
|
|
47
|
+
def build_high_dim_probabilities(
|
|
48
|
+
mol=None,
|
|
49
|
+
dist: Optional[np.ndarray] = None,
|
|
50
|
+
dist_name: str = "D2",
|
|
51
|
+
compute_D_params: Optional[Dict[str, Any]] = None,
|
|
52
|
+
hop: Optional[np.ndarray] = None,
|
|
53
|
+
K_HOP_MAX: int = 2,
|
|
54
|
+
k_min: int = 3,
|
|
55
|
+
k_max: int = 8,
|
|
56
|
+
default_k_when_no_deg: float = 6.0,
|
|
57
|
+
sigma_search_iters: int = 32,
|
|
58
|
+
sigma_hi_init: float = 1.0,
|
|
59
|
+
verbose: bool = False,
|
|
60
|
+
data=None,
|
|
61
|
+
):
|
|
62
|
+
"""
|
|
63
|
+
构造高维概率图(返回对称化的 P 矩阵以及中间产物)。
|
|
64
|
+
|
|
65
|
+
返回 dict:
|
|
66
|
+
- "P": 对称化的概率矩阵 (n x n)
|
|
67
|
+
- "prob": 非对称未归一化的 prob 矩阵
|
|
68
|
+
- "sigma": 每行二分搜索得到的 sigma
|
|
69
|
+
- "dist": 使用的距离矩阵 (n x n)
|
|
70
|
+
- "hop": 使用的 hop 矩阵或 None
|
|
71
|
+
"""
|
|
72
|
+
|
|
73
|
+
params = compute_D_params or {}
|
|
74
|
+
|
|
75
|
+
# ---------- 1. 准备距离矩阵 ----------
|
|
76
|
+
if dist is None:
|
|
77
|
+
if mol is None:
|
|
78
|
+
raise ValueError("必须提供 dist 矩阵,或提供 mol 以便根据 dist_name 计算。")
|
|
79
|
+
|
|
80
|
+
if verbose:
|
|
81
|
+
print(f"[build_high_dim_probabilities] computing distance {dist_name} ...")
|
|
82
|
+
|
|
83
|
+
if dist_name == "D1":
|
|
84
|
+
raw = compute_augmented_graph_distance_np(mol=mol, data=data, **params)
|
|
85
|
+
dist_mat = _extract_distance_matrix(raw, prefer_key="D_aug")
|
|
86
|
+
elif dist_name == "D2":
|
|
87
|
+
raw = compute_AE_tanimoto_distance_np(mol=mol, **params)
|
|
88
|
+
# AE / tanimoto 函数通常返回矩阵,但也可能返回 dict
|
|
89
|
+
dist_mat = _extract_distance_matrix(raw, prefer_key=None)
|
|
90
|
+
elif dist_name == "D3":
|
|
91
|
+
raw = compute_embed3d_distance_np(mol=mol, **params)
|
|
92
|
+
dist_mat = _extract_distance_matrix(raw, prefer_key=None)
|
|
93
|
+
else:
|
|
94
|
+
raise ValueError(f"未知 dist_name: {dist_name}")
|
|
95
|
+
else:
|
|
96
|
+
dist_mat = np.asarray(dist, dtype=float)
|
|
97
|
+
if dist_mat.ndim != 2 or dist_mat.shape[0] != dist_mat.shape[1]:
|
|
98
|
+
raise ValueError("传入的 dist 必须是方阵 (n x n). got: %s" % (dist_mat.shape,))
|
|
99
|
+
|
|
100
|
+
n = int(dist_mat.shape[0])
|
|
101
|
+
|
|
102
|
+
if verbose:
|
|
103
|
+
print(f"[build_high_dim_probabilities] distance matrix shape: {dist_mat.shape}")
|
|
104
|
+
|
|
105
|
+
# ---------- 2. 准备 hop 矩阵 ----------
|
|
106
|
+
hop_used = hop
|
|
107
|
+
if hop_used is None and mol is not None:
|
|
108
|
+
try:
|
|
109
|
+
hop_used = hop_matrix_from_mol(mol)
|
|
110
|
+
except Exception as e:
|
|
111
|
+
if verbose:
|
|
112
|
+
print(f"[build_high_dim_probabilities] hop_matrix_from_mol failed: {e}. Ignoring hop.")
|
|
113
|
+
hop_used = None
|
|
114
|
+
|
|
115
|
+
if hop_used is not None:
|
|
116
|
+
hop_used = np.asarray(hop_used, dtype=float)
|
|
117
|
+
if hop_used.shape != (n, n):
|
|
118
|
+
if verbose:
|
|
119
|
+
print("[build_high_dim_probabilities] Warning: hop 矩阵形状与 dist 不匹配,忽略 hop 筛选。")
|
|
120
|
+
hop_used = None
|
|
121
|
+
|
|
122
|
+
# ---------- 3. 计算每个节点目标 k_i ----------
|
|
123
|
+
try:
|
|
124
|
+
if mol is not None:
|
|
125
|
+
deg = np.array([a.GetDegree() for a in mol.GetAtoms()], dtype=int)
|
|
126
|
+
if deg.shape[0] != n:
|
|
127
|
+
# 不匹配则忽略 deg
|
|
128
|
+
deg = None
|
|
129
|
+
else:
|
|
130
|
+
deg = None
|
|
131
|
+
|
|
132
|
+
if deg is not None:
|
|
133
|
+
k_i = np.clip(deg + 2, k_min, k_max).astype(float)
|
|
134
|
+
else:
|
|
135
|
+
k_i = np.full(n, default_k_when_no_deg, dtype=float)
|
|
136
|
+
except Exception:
|
|
137
|
+
k_i = np.full(n, default_k_when_no_deg, dtype=float)
|
|
138
|
+
|
|
139
|
+
# ---------- 4. 辅助函数 ----------
|
|
140
|
+
def prob_row_from_sigma(sigma, d_row, rho_i):
|
|
141
|
+
x = d_row - rho_i
|
|
142
|
+
x = np.where(x < 0.0, 0.0, x)
|
|
143
|
+
return np.exp(- x / (sigma + 1e-12))
|
|
144
|
+
|
|
145
|
+
def k_from_prob(prob_row):
|
|
146
|
+
s = float(prob_row.sum())
|
|
147
|
+
return np.power(2.0, s) - 1.0
|
|
148
|
+
|
|
149
|
+
def sigma_binary_search_row(k_of_sigma, fixed_k, lo=1e-6, hi=sigma_hi_init, iters=sigma_search_iters):
|
|
150
|
+
# expand hi until >= fixed_k (robust策略)
|
|
151
|
+
try:
|
|
152
|
+
while k_of_sigma(hi) < fixed_k:
|
|
153
|
+
hi *= 2.0
|
|
154
|
+
if hi > 1e6:
|
|
155
|
+
break
|
|
156
|
+
except Exception:
|
|
157
|
+
hi = sigma_hi_init
|
|
158
|
+
|
|
159
|
+
for _ in range(iters):
|
|
160
|
+
mid = 0.5 * (lo + hi)
|
|
161
|
+
try:
|
|
162
|
+
if k_of_sigma(mid) < fixed_k:
|
|
163
|
+
lo = mid
|
|
164
|
+
else:
|
|
165
|
+
hi = mid
|
|
166
|
+
except Exception:
|
|
167
|
+
lo = mid
|
|
168
|
+
hi *= 2.0
|
|
169
|
+
return 0.5 * (lo + hi)
|
|
170
|
+
|
|
171
|
+
# ---------- 5. 主循环 ----------
|
|
172
|
+
prob = np.zeros((n, n), dtype=float)
|
|
173
|
+
sigma_array = np.zeros(n, dtype=float)
|
|
174
|
+
|
|
175
|
+
for i in range(n):
|
|
176
|
+
# 全量可选集合:非自身且距离有限
|
|
177
|
+
finite_mask = np.isfinite(dist_mat[i])
|
|
178
|
+
cand = np.where((np.arange(n) != i) & finite_mask)[0]
|
|
179
|
+
|
|
180
|
+
# 若有 hop 限制,则再筛一次
|
|
181
|
+
if hop_used is not None:
|
|
182
|
+
hop_mask = hop_used[i] <= K_HOP_MAX
|
|
183
|
+
cand = cand[hop_mask[cand]]
|
|
184
|
+
|
|
185
|
+
if cand.size == 0:
|
|
186
|
+
sigma_array[i] = 0.0
|
|
187
|
+
if verbose and ((i + 1) % 200 == 0):
|
|
188
|
+
print(f"Row {i}: no candidates (all inf or excluded by hop)")
|
|
189
|
+
continue
|
|
190
|
+
|
|
191
|
+
if cand.size <= 2:
|
|
192
|
+
prob[i, cand] = 1.0 / float(cand.size)
|
|
193
|
+
sigma_array[i] = 0.0
|
|
194
|
+
continue
|
|
195
|
+
|
|
196
|
+
d_i = dist_mat[i, cand].astype(float)
|
|
197
|
+
|
|
198
|
+
pos = d_i[d_i > 0.0]
|
|
199
|
+
rho_i = float(pos.min()) if pos.size > 0 else 0.0
|
|
200
|
+
|
|
201
|
+
k_target = float(min(k_i[i], cand.size))
|
|
202
|
+
|
|
203
|
+
def k_of_sigma(sig):
|
|
204
|
+
p = prob_row_from_sigma(sig, d_i, rho_i)
|
|
205
|
+
return k_from_prob(p)
|
|
206
|
+
|
|
207
|
+
sigma_i = sigma_binary_search_row(k_of_sigma, k_target, lo=1e-6, hi=sigma_hi_init, iters=sigma_search_iters)
|
|
208
|
+
sigma_array[i] = sigma_i
|
|
209
|
+
|
|
210
|
+
p_row = prob_row_from_sigma(sigma_i, d_i, rho_i)
|
|
211
|
+
prob[i, cand] = p_row
|
|
212
|
+
|
|
213
|
+
if verbose and ((i + 1) % 100 == 0 or i == n - 1):
|
|
214
|
+
print(f"Sigma search row {i+1}/{n}")
|
|
215
|
+
|
|
216
|
+
if verbose:
|
|
217
|
+
print("Mean sigma =", float(sigma_array.mean()))
|
|
218
|
+
|
|
219
|
+
# ---------- 6. 对称化(fuzzy union) ----------
|
|
220
|
+
P = prob + prob.T - prob * prob.T
|
|
221
|
+
|
|
222
|
+
return {
|
|
223
|
+
"P": P,
|
|
224
|
+
# "prob": prob,
|
|
225
|
+
# "sigma": sigma_array,
|
|
226
|
+
# "dist": dist_mat,
|
|
227
|
+
# "hop": hop_used,
|
|
228
|
+
}
|
|
229
|
+
|
|
230
|
+
|
|
231
|
+
# ---------------- 简要示例 ----------------
|
|
232
|
+
# result = build_high_dim_probabilities(mol=mol, dist_name="D2", compute_D_params={'radius':2,'nBits':2048})
|
|
233
|
+
# P = result['P']
|
|
234
|
+
|
|
235
|
+
|
|
236
|
+
|
|
237
|
+
|
|
@@ -0,0 +1,124 @@
|
|
|
1
|
+
"""Lazy, shard-backed reader for the isolated full top20 experiments.
|
|
2
|
+
|
|
3
|
+
The legacy InMemoryDataset remains the default. This reader only loads the
|
|
4
|
+
small shard containing a requested sample and never materializes RDKit objects.
|
|
5
|
+
"""
|
|
6
|
+
from __future__ import annotations
|
|
7
|
+
|
|
8
|
+
import json
|
|
9
|
+
import os
|
|
10
|
+
from bisect import bisect_right
|
|
11
|
+
|
|
12
|
+
import torch
|
|
13
|
+
from torch_geometric.data import Data
|
|
14
|
+
from torch.utils.data import Subset
|
|
15
|
+
|
|
16
|
+
|
|
17
|
+
class ShardedDrugDataset:
|
|
18
|
+
def __init__(self, root, split="train", cache_size=2, assignment_root=""):
|
|
19
|
+
self.root = os.fspath(root)
|
|
20
|
+
with open(os.path.join(self.root, "manifest.json"), "r") as handle:
|
|
21
|
+
self.manifest = json.load(handle)
|
|
22
|
+
if split not in self.manifest["splits"]:
|
|
23
|
+
raise ValueError("Unknown shard split: %s" % split)
|
|
24
|
+
info = self.manifest["splits"][split]
|
|
25
|
+
self.split = split
|
|
26
|
+
self.shards = info["shards"]
|
|
27
|
+
self.offsets = [int(item["start"]) for item in self.shards]
|
|
28
|
+
self.length = int(info["samples"])
|
|
29
|
+
self.molecules = int(info["molecules"])
|
|
30
|
+
self._cache_size = max(1, int(cache_size))
|
|
31
|
+
self._cache = {}
|
|
32
|
+
self._cache_order = []
|
|
33
|
+
self._packed = self.manifest.get("format") == "manimol-shards-v3-packed"
|
|
34
|
+
self._assignments = {}
|
|
35
|
+
if assignment_root:
|
|
36
|
+
assignment_path = os.path.join(os.fspath(assignment_root), "%s.pt" % split)
|
|
37
|
+
if os.path.exists(assignment_path):
|
|
38
|
+
assignment_payload = torch.load(
|
|
39
|
+
assignment_path, map_location="cpu", weights_only=False
|
|
40
|
+
)
|
|
41
|
+
source_identity = assignment_payload.get("source_identity", "")
|
|
42
|
+
expected_identity = self.manifest.get("identity", {}).get("sha256", "")
|
|
43
|
+
if source_identity and expected_identity and source_identity != expected_identity:
|
|
44
|
+
raise ValueError("Stage-II assignment cache does not match shard identity")
|
|
45
|
+
self._assignments = assignment_payload.get("molecules", {})
|
|
46
|
+
self.molecule_groups_by_shard = []
|
|
47
|
+
self._group_ends_by_shard = []
|
|
48
|
+
for shard in self.shards:
|
|
49
|
+
cursor = int(shard["start"])
|
|
50
|
+
groups, ends, local_total = [], [], 0
|
|
51
|
+
for size in shard.get("group_sizes", []):
|
|
52
|
+
size = int(size)
|
|
53
|
+
groups.append(range(cursor, cursor + size))
|
|
54
|
+
cursor += size
|
|
55
|
+
local_total += size
|
|
56
|
+
ends.append(local_total)
|
|
57
|
+
if groups:
|
|
58
|
+
if cursor != int(shard["start"]) + int(shard["samples"]):
|
|
59
|
+
raise ValueError("Invalid molecule group metadata in %s" % shard["file"])
|
|
60
|
+
self.molecule_groups_by_shard.append(groups)
|
|
61
|
+
self._group_ends_by_shard.append(ends)
|
|
62
|
+
if len(self.molecule_groups_by_shard) != len(self.shards):
|
|
63
|
+
self.molecule_groups_by_shard = []
|
|
64
|
+
self._group_ends_by_shard = []
|
|
65
|
+
|
|
66
|
+
def __len__(self):
|
|
67
|
+
return self.length
|
|
68
|
+
|
|
69
|
+
def _load_shard(self, shard_id):
|
|
70
|
+
shard_id = int(shard_id)
|
|
71
|
+
if shard_id in self._cache:
|
|
72
|
+
return self._cache[shard_id]
|
|
73
|
+
path = os.path.join(self.root, self.shards[shard_id]["file"])
|
|
74
|
+
payload = torch.load(path, map_location="cpu", weights_only=False)
|
|
75
|
+
records = payload["molecules"] if self._packed else payload["records"]
|
|
76
|
+
self._cache[shard_id] = records
|
|
77
|
+
self._cache_order.append(shard_id)
|
|
78
|
+
while len(self._cache_order) > self._cache_size:
|
|
79
|
+
old = self._cache_order.pop(0)
|
|
80
|
+
self._cache.pop(old, None)
|
|
81
|
+
return records
|
|
82
|
+
|
|
83
|
+
def __getitem__(self, index):
|
|
84
|
+
if isinstance(index, (list, tuple, torch.Tensor)):
|
|
85
|
+
if isinstance(index, torch.Tensor):
|
|
86
|
+
index = index.tolist()
|
|
87
|
+
return Subset(self, list(index))
|
|
88
|
+
index = int(index)
|
|
89
|
+
if index < 0:
|
|
90
|
+
index += self.length
|
|
91
|
+
if index < 0 or index >= self.length:
|
|
92
|
+
raise IndexError(index)
|
|
93
|
+
shard_id = bisect_right(self.offsets, index) - 1
|
|
94
|
+
local_index = index - self.offsets[shard_id]
|
|
95
|
+
records = self._load_shard(shard_id)
|
|
96
|
+
if not self._packed:
|
|
97
|
+
return records[local_index]
|
|
98
|
+
ends = self._group_ends_by_shard[shard_id]
|
|
99
|
+
molecule_index = bisect_right(ends, local_index)
|
|
100
|
+
start = 0 if molecule_index == 0 else ends[molecule_index - 1]
|
|
101
|
+
conformer_index = local_index - start
|
|
102
|
+
molecule = records[molecule_index]
|
|
103
|
+
fields = {key: value for key, value in molecule.items()
|
|
104
|
+
if key not in ("pos", "conf_id")}
|
|
105
|
+
fields["pos"] = molecule["pos"][conformer_index]
|
|
106
|
+
fields["conf_id"] = molecule["conf_id"][conformer_index]
|
|
107
|
+
mol_id = int(molecule["mol_id"].item())
|
|
108
|
+
assignment = self._assignments.get(mol_id)
|
|
109
|
+
if assignment is not None:
|
|
110
|
+
expected_conf = int(assignment["conf_id"][conformer_index].item())
|
|
111
|
+
actual_conf = int(fields["conf_id"].item())
|
|
112
|
+
if expected_conf != actual_conf:
|
|
113
|
+
raise ValueError(
|
|
114
|
+
"Assignment conf_id mismatch for molecule %d: %d != %d"
|
|
115
|
+
% (mol_id, expected_conf, actual_conf)
|
|
116
|
+
)
|
|
117
|
+
fields["stage2_cluster_id"] = assignment["cluster_id"][conformer_index]
|
|
118
|
+
fields["stage2_cluster_count"] = torch.tensor(
|
|
119
|
+
int(assignment["cluster_count"]), dtype=torch.long
|
|
120
|
+
)
|
|
121
|
+
return Data(**fields)
|
|
122
|
+
|
|
123
|
+
def split_indices(self):
|
|
124
|
+
return list(range(self.length))
|
dataset/smiles2graph.py
ADDED
|
@@ -0,0 +1,217 @@
|
|
|
1
|
+
# import dgl
|
|
2
|
+
import numpy as np
|
|
3
|
+
import rdkit
|
|
4
|
+
import torch
|
|
5
|
+
from rdkit import Chem
|
|
6
|
+
import os
|
|
7
|
+
from rdkit.Chem import AllChem
|
|
8
|
+
from models import (
|
|
9
|
+
KERNEL, UMAPLowKernel, GaussianKernel, StudentTKernel, pairwise_dist,
|
|
10
|
+
UMAPRowExpKernel, UMAPRowFamilyKernel, SmoothKRowExpKernel, find_ab_params,
|
|
11
|
+
)
|
|
12
|
+
from models import KERNEL, pairwise_dist
|
|
13
|
+
|
|
14
|
+
|
|
15
|
+
def get_atom_features(atom):
|
|
16
|
+
# The usage of features is along with the Attentive FP.
|
|
17
|
+
feature = np.zeros(39)
|
|
18
|
+
|
|
19
|
+
# Symbol
|
|
20
|
+
symbol = atom.GetSymbol()
|
|
21
|
+
symbol_list = ['H', 'B', 'C', 'N', 'O', 'F', 'Si', 'P', 'S', 'Cl', 'As', 'Se', 'Br', 'I', 'At']
|
|
22
|
+
if symbol in symbol_list:
|
|
23
|
+
loc = symbol_list.index(symbol)
|
|
24
|
+
feature[loc] = 1
|
|
25
|
+
else:
|
|
26
|
+
feature[15] = 1
|
|
27
|
+
|
|
28
|
+
# Degree
|
|
29
|
+
degree = atom.GetDegree()
|
|
30
|
+
if degree > 5:
|
|
31
|
+
print("atom degree larger than 5. Please check before featurizing.")
|
|
32
|
+
raise RuntimeError
|
|
33
|
+
|
|
34
|
+
feature[16 + degree] = 1
|
|
35
|
+
|
|
36
|
+
# Formal Charge
|
|
37
|
+
charge = atom.GetFormalCharge()
|
|
38
|
+
feature[22] = charge
|
|
39
|
+
|
|
40
|
+
# radical electrons
|
|
41
|
+
radelc = atom.GetNumRadicalElectrons()
|
|
42
|
+
feature[23] = radelc
|
|
43
|
+
|
|
44
|
+
# Hybridization
|
|
45
|
+
hyb = atom.GetHybridization()
|
|
46
|
+
hybridization_list = [rdkit.Chem.rdchem.HybridizationType.SP,
|
|
47
|
+
rdkit.Chem.rdchem.HybridizationType.SP2,
|
|
48
|
+
rdkit.Chem.rdchem.HybridizationType.SP3,
|
|
49
|
+
rdkit.Chem.rdchem.HybridizationType.SP3D,
|
|
50
|
+
rdkit.Chem.rdchem.HybridizationType.SP3D2]
|
|
51
|
+
if hyb in hybridization_list:
|
|
52
|
+
loc = hybridization_list.index(hyb)
|
|
53
|
+
feature[loc + 24] = 1
|
|
54
|
+
else:
|
|
55
|
+
feature[29] = 1
|
|
56
|
+
|
|
57
|
+
# aromaticity
|
|
58
|
+
if atom.GetIsAromatic():
|
|
59
|
+
feature[30] = 1
|
|
60
|
+
|
|
61
|
+
# hydrogens
|
|
62
|
+
hs = atom.GetNumImplicitHs()
|
|
63
|
+
feature[31 + hs] = 1
|
|
64
|
+
|
|
65
|
+
# chirality, chirality type
|
|
66
|
+
if atom.HasProp('_ChiralityPossible'):
|
|
67
|
+
# TODO what kind of error
|
|
68
|
+
feature[36] = 1
|
|
69
|
+
|
|
70
|
+
try:
|
|
71
|
+
chi = atom.GetProp('_CIPCode')
|
|
72
|
+
chi_list = ['R', 'S']
|
|
73
|
+
loc = chi_list.index(chi)
|
|
74
|
+
feature[37 + loc] = 1
|
|
75
|
+
except KeyError:
|
|
76
|
+
feature[37] = 0
|
|
77
|
+
feature[38] = 0
|
|
78
|
+
|
|
79
|
+
return feature
|
|
80
|
+
|
|
81
|
+
|
|
82
|
+
def get_bond_features(bond):
|
|
83
|
+
feature = np.zeros(10)
|
|
84
|
+
|
|
85
|
+
# bond type
|
|
86
|
+
type = bond.GetBondType()
|
|
87
|
+
bond_type_list = [rdkit.Chem.rdchem.BondType.SINGLE,
|
|
88
|
+
rdkit.Chem.rdchem.BondType.DOUBLE,
|
|
89
|
+
rdkit.Chem.rdchem.BondType.TRIPLE,
|
|
90
|
+
rdkit.Chem.rdchem.BondType.AROMATIC]
|
|
91
|
+
if type in bond_type_list:
|
|
92
|
+
loc = bond_type_list.index(type)
|
|
93
|
+
feature[0 + loc] = 1
|
|
94
|
+
else:
|
|
95
|
+
print("Wrong type of bond. Please check before feturization.")
|
|
96
|
+
raise RuntimeError
|
|
97
|
+
|
|
98
|
+
# conjugation
|
|
99
|
+
conj = bond.GetIsConjugated()
|
|
100
|
+
feature[4] = conj
|
|
101
|
+
|
|
102
|
+
# ring
|
|
103
|
+
ring = bond.IsInRing()
|
|
104
|
+
feature[5] = ring
|
|
105
|
+
|
|
106
|
+
# stereo
|
|
107
|
+
stereo = bond.GetStereo()
|
|
108
|
+
stereo_list = [rdkit.Chem.rdchem.BondStereo.STEREONONE,
|
|
109
|
+
rdkit.Chem.rdchem.BondStereo.STEREOANY,
|
|
110
|
+
rdkit.Chem.rdchem.BondStereo.STEREOZ,
|
|
111
|
+
rdkit.Chem.rdchem.BondStereo.STEREOE]
|
|
112
|
+
if stereo in stereo_list:
|
|
113
|
+
loc = stereo_list.index(stereo)
|
|
114
|
+
feature[6 + loc] = 1
|
|
115
|
+
else:
|
|
116
|
+
print("Wrong stereo type of bond. Please check before featurization.")
|
|
117
|
+
raise RuntimeError
|
|
118
|
+
|
|
119
|
+
return feature
|
|
120
|
+
|
|
121
|
+
|
|
122
|
+
def get_atom_vdw_radii(mol):
|
|
123
|
+
"""计算每个原子的范德华半径"""
|
|
124
|
+
periodic_table = rdkit.Chem.GetPeriodicTable()
|
|
125
|
+
vdw_radii = []
|
|
126
|
+
for atom in mol.GetAtoms():
|
|
127
|
+
atomic_num = atom.GetAtomicNum()
|
|
128
|
+
# 获取范德华半径 (单位: Angstrom)
|
|
129
|
+
radius = periodic_table.GetRvdw(atomic_num)
|
|
130
|
+
vdw_radii.append(radius)
|
|
131
|
+
return torch.tensor(vdw_radii, dtype=torch.float).view(-1, 1)
|
|
132
|
+
|
|
133
|
+
|
|
134
|
+
def smile2graph4GEOM(data):
|
|
135
|
+
mol = data['rdmol']
|
|
136
|
+
# if (mol is None):
|
|
137
|
+
# return None
|
|
138
|
+
src = []
|
|
139
|
+
dst = []
|
|
140
|
+
atom_feature = []
|
|
141
|
+
bond_feature = []
|
|
142
|
+
|
|
143
|
+
for atom in mol.GetAtoms():
|
|
144
|
+
one_atom_feature = get_atom_features(atom)
|
|
145
|
+
atom_feature.append(one_atom_feature)
|
|
146
|
+
atom_feature = np.array(atom_feature)
|
|
147
|
+
atom_feature = torch.tensor(atom_feature).float()
|
|
148
|
+
vdw_radii = get_atom_vdw_radii(mol)
|
|
149
|
+
if len(mol.GetBonds()) > 0: # mol has bonds
|
|
150
|
+
for bond in mol.GetBonds():
|
|
151
|
+
i = bond.GetBeginAtomIdx()
|
|
152
|
+
j = bond.GetEndAtomIdx()
|
|
153
|
+
one_bond_feature = get_bond_features(bond)
|
|
154
|
+
src.append(i)
|
|
155
|
+
dst.append(j)
|
|
156
|
+
bond_feature.append(one_bond_feature)
|
|
157
|
+
src.append(j)
|
|
158
|
+
dst.append(i)
|
|
159
|
+
bond_feature.append(one_bond_feature)
|
|
160
|
+
|
|
161
|
+
src = torch.tensor(src).long()
|
|
162
|
+
dst = torch.tensor(dst).long()
|
|
163
|
+
bond_feature = np.array(bond_feature)
|
|
164
|
+
bond_feature = torch.tensor(bond_feature).float()
|
|
165
|
+
edge_index = torch.vstack([src, dst])
|
|
166
|
+
# graph_cur_smile = dgl.graph((src, dst), num_nodes=len(mol.GetAtoms()))
|
|
167
|
+
# graph_cur_smile.ndata['x'] = atom_feature
|
|
168
|
+
# graph_cur_smile.edata['x'] = bond_feature
|
|
169
|
+
else:
|
|
170
|
+
edge_index = torch.empty((2, 0)).long()
|
|
171
|
+
bond_feature = torch.empty((0, 10)).float()
|
|
172
|
+
|
|
173
|
+
return atom_feature, edge_index, bond_feature, vdw_radii
|
|
174
|
+
|
|
175
|
+
|
|
176
|
+
def teacher_coords_from_smiles(mol, seed=42, optimize=True):
|
|
177
|
+
"""
|
|
178
|
+
用 RDKit ETKDG 生成 3D 坐标(仅重原子,保持与你可视化的原子顺序一致)。
|
|
179
|
+
如果优化=True,做一次 UFF 优化以稳定键长。
|
|
180
|
+
失败时返回 (None, None)。
|
|
181
|
+
"""
|
|
182
|
+
mol_copy = Chem.Mol(mol) # 避免修改原对象
|
|
183
|
+
# 尝试多次嵌入(有时随机种子影响成功)
|
|
184
|
+
for attempt in range(3):
|
|
185
|
+
try:
|
|
186
|
+
params = AllChem.ETKDGv3()
|
|
187
|
+
params.randomSeed = seed + attempt
|
|
188
|
+
params.useSmallRingTorsions = True
|
|
189
|
+
ok = AllChem.EmbedMolecule(mol_copy, params)
|
|
190
|
+
if ok == 0:
|
|
191
|
+
if optimize:
|
|
192
|
+
try:
|
|
193
|
+
AllChem.MMFFOptimizeMolecule(mol_copy, maxIters=1000)
|
|
194
|
+
except Exception:
|
|
195
|
+
pass
|
|
196
|
+
conf = mol_copy.GetConformer(0)
|
|
197
|
+
idx_map = [conf.GetAtomPosition(a.GetIdx()) for a in mol_copy.GetAtoms()]
|
|
198
|
+
Y_true = np.array([[p.x, p.y, p.z] for p in idx_map], dtype=float)
|
|
199
|
+
return mol_copy, Y_true
|
|
200
|
+
except Exception:
|
|
201
|
+
continue
|
|
202
|
+
return None, None
|
|
203
|
+
|
|
204
|
+
|
|
205
|
+
def center_and_rescale(Y, target_rms=1.0):
|
|
206
|
+
"""每个 epoch:居中并把坐标 RMS 缩放到 target_rms,防止数值漂移或坍缩。"""
|
|
207
|
+
Y = Y - Y.mean(axis=0, keepdims=True)
|
|
208
|
+
rms = np.sqrt((Y ** 2).mean())
|
|
209
|
+
if rms < 1e-8:
|
|
210
|
+
return Y
|
|
211
|
+
return Y * (target_rms / rms)
|
|
212
|
+
|
|
213
|
+
|
|
214
|
+
def q_from_Y(Y: np.ndarray, KERNEL):
|
|
215
|
+
D = pairwise_dist(Y)
|
|
216
|
+
Q = KERNEL.forward(D)
|
|
217
|
+
return Q, D
|
dualmapref/__init__.py
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
1
|
+
"""Public DualMapRef API, backed by the checkpoint-compatible implementation."""
|
|
2
|
+
from conformend import (
|
|
3
|
+
ConforMend as _LegacyConforMend,
|
|
4
|
+
circular_standard_deviation,
|
|
5
|
+
density_centrality_select,
|
|
6
|
+
ensemble_pairwise_prior,
|
|
7
|
+
fps_indices,
|
|
8
|
+
mmff94s_window_fps,
|
|
9
|
+
pairwise_rmsd,
|
|
10
|
+
run_full_inference,
|
|
11
|
+
select_energy_window,
|
|
12
|
+
wrap_angle,
|
|
13
|
+
)
|
|
14
|
+
|
|
15
|
+
class DualMapRef(_LegacyConforMend):
|
|
16
|
+
"""Checkpoint-backed DualMapRef inference interface."""
|
|
17
|
+
|
|
18
|
+
|
|
19
|
+
__all__ = [
|
|
20
|
+
'DualMapRef', 'circular_standard_deviation', 'density_centrality_select',
|
|
21
|
+
'ensemble_pairwise_prior', 'fps_indices', 'mmff94s_window_fps',
|
|
22
|
+
'pairwise_rmsd', 'run_full_inference', 'select_energy_window', 'wrap_angle',
|
|
23
|
+
]
|
|
24
|
+
__version__ = '0.2.8'
|
dualmapref/refiner.py
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
1
|
+
"""DualMapRef candidate-conditioned torsional refinement API."""
|
|
2
|
+
from conformend.refiner import (
|
|
3
|
+
AcceptanceConfig, ConforMendRefiner as _LegacyRefiner,
|
|
4
|
+
InterpretableLocalRefiner, RefinementResult, accept_refinement,
|
|
5
|
+
apply_torsion_residuals, bond_cross_fragment_context, proximity_matrix,
|
|
6
|
+
stable_relational_energy,
|
|
7
|
+
)
|
|
8
|
+
class DualMapRefRefiner(_LegacyRefiner):
|
|
9
|
+
"""Checkpoint-compatible DualMapRef torsional Refiner."""
|
|
10
|
+
|
|
11
|
+
|
|
12
|
+
__all__ = [
|
|
13
|
+
'AcceptanceConfig', 'DualMapRefRefiner', 'InterpretableLocalRefiner',
|
|
14
|
+
'RefinementResult', 'accept_refinement', 'apply_torsion_residuals',
|
|
15
|
+
'bond_cross_fragment_context', 'proximity_matrix', 'stable_relational_energy',
|
|
16
|
+
]
|
|
@@ -0,0 +1,21 @@
|
|
|
1
|
+
MIT License
|
|
2
|
+
|
|
3
|
+
Copyright (c) 2026 ConforMend developers
|
|
4
|
+
|
|
5
|
+
Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
6
|
+
of this software and associated documentation files (the "Software"), to deal
|
|
7
|
+
in the Software without restriction, including without limitation the rights
|
|
8
|
+
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
|
9
|
+
copies of the Software, and to permit persons to whom the Software is
|
|
10
|
+
furnished to do so, subject to the following conditions:
|
|
11
|
+
|
|
12
|
+
The above copyright notice and this permission notice shall be included in all
|
|
13
|
+
copies or substantial portions of the Software.
|
|
14
|
+
|
|
15
|
+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
16
|
+
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
17
|
+
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
|
18
|
+
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
19
|
+
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
20
|
+
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
|
21
|
+
SOFTWARE.
|