Hani Park commited on
Commit
2fd49e8
·
1 Parent(s): 01f16f6

Removed invalid SMILES and updated README.md

Browse files
Files changed (1) hide show
  1. README.md +167 -26
README.md CHANGED
@@ -1,4 +1,43 @@
1
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2
  dataset_info:
3
  - config_name: Molecule3D_random_split
4
  features:
@@ -25,15 +64,15 @@ dataset_info:
25
  - name: scf energy
26
  dtype: float64
27
  splits:
28
- - name: train
29
- num_bytes: 3175756092
30
- num_examples: 2339728
31
- - name: test
32
- num_bytes: 1058783091
33
- num_examples: 779895
34
- - name: validation
35
- num_bytes: 1058496510
36
- num_examples: 779903
37
  download_size: 1881829100
38
  dataset_size: 5293035693
39
  - config_name: Molecule3D_scaffold_split
@@ -72,21 +111,123 @@ dataset_info:
72
  num_examples: 779859
73
  download_size: 1867676945
74
  dataset_size: 5293035693
75
- configs:
76
- - config_name: Molecule3D_random_split
77
- data_files:
78
- - split: train
79
- path: Molecule3D_random_split/train-*
80
- - split: test
81
- path: Molecule3D_random_split/test-*
82
- - split: validation
83
- path: Molecule3D_random_split/validation-*
84
- - config_name: Molecule3D_scaffold_split
85
- data_files:
86
- - split: train
87
- path: Molecule3D_scaffold_split/train-*
88
- - split: validation
89
- path: Molecule3D_scaffold_split/validation-*
90
- - split: test
91
- path: Molecule3D_scaffold_split/test-*
92
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ version: 1.1.1
3
+ language: en
4
+ license: gpl-3.0
5
+ size_categories:
6
+ - 1M<n<10M
7
+ task_categories:
8
+ - tabular-regression
9
+ pretty_name: Molecule3D
10
+ tags:
11
+ - molecular geometry
12
+ - molecular graph
13
+ dataset_summary: Curated dataset of ground-state geometries of 4 million molecules
14
+ dervied from density functional theory, consisting of SMILES, sdf, and 3D properties
15
+ of molecules. Random split and scaffold split datasets are uploaded to our repository.
16
+ citation: '@misc{https://doi.org/10.48550/arxiv.2110.01717, doi = {10.48550/ARXIV.2110.01717},
17
+ url = {https://arxiv.org/abs/2110.01717}, author = {Xu, Zhao and Luo, Youzhi and
18
+ Zhang, Xuan and Xu, Xinyi and Xie, Yaochen and Liu, Meng and Dickerson, Kaleb
19
+ and Deng, Cheng and Nakata, Maho and Ji, Shuiwang}, keywords = {Machine Learning
20
+ (cs.LG), Artificial Intelligence (cs.AI), FOS: Computer and information sciences, FOS:
21
+ Computer and information sciences}, title = {Molecule3D: A Benchmark for Predicting
22
+ 3D Geometries from Molecular Graphs}, publisher = {arXiv}, year = {2021}, copyright
23
+ = {arXiv.org perpetual, non-exclusive license} }'
24
+ configs:
25
+ - config_name: Molecule3D_random_split
26
+ data_files:
27
+ - split: train
28
+ path: Molecule3D_random_split/train-*
29
+ - split: test
30
+ path: Molecule3D_random_split/test-*
31
+ - split: validation
32
+ path: Molecule3D_random_split/validation-*
33
+ - config_name: Molecule3D_scaffold_split
34
+ data_files:
35
+ - split: train
36
+ path: Molecule3D_scaffold_split/train-*
37
+ - split: validation
38
+ path: Molecule3D_scaffold_split/validation-*
39
+ - split: test
40
+ path: Molecule3D_scaffold_split/test-*
41
  dataset_info:
42
  - config_name: Molecule3D_random_split
43
  features:
 
64
  - name: scf energy
65
  dtype: float64
66
  splits:
67
+ - name: train
68
+ num_bytes: 3175756092
69
+ num_examples: 2339728
70
+ - name: test
71
+ num_bytes: 1058783091
72
+ num_examples: 779895
73
+ - name: validation
74
+ num_bytes: 1058496510
75
+ num_examples: 779903
76
  download_size: 1881829100
77
  dataset_size: 5293035693
78
  - config_name: Molecule3D_scaffold_split
 
111
  num_examples: 779859
112
  download_size: 1867676945
113
  dataset_size: 5293035693
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
114
  ---
115
+
116
+ # Molecule3D
117
+ [Molecule3D](https://arxiv.org/abs/2110.01717) is a comprehensive dataset containing ground-state geometries derived from Density Functional Theory (DFT) calculations for approximately 4 million molecules.
118
+ This is a mirror of the [Official Github repo](https://github.com/divelab/MoleculeX/tree/molx/Molecule3D) where the dataset was uploaded in 2021.
119
+
120
+
121
+ ## Preprocseeing
122
+ [Update on 2025.08.16 - version 1.1.1]
123
+ We removed invalid SMILES strings which could not be parsed by RDKit.
124
+ - Random split
125
+ 1. train : removed 60 smiles strings from 2339788 strings
126
+ 2. test : removed 35 smiles strings from 779930 strings
127
+ 3. validation : removed 26 smiles strings from 779929 strings
128
+ - Scaffold split
129
+ 1. train : removed 46 smiles strings from 2339788 strings
130
+ 2. test : removed 71 smiles strings from 779930 strings
131
+ 3. validation : removed 4 smiles strings from 779929 strings
132
+
133
+ We also updated the `README.md` file.
134
+
135
+ We utilized the raw data uploaded on [Github](https://github.com/divelab/MoleculeX/tree/molx/Molecule3D/data/raw) and performed several preprocessing:
136
+ 1. Sanitize the molecules using RDKit and MolVS (standardize SMILES format)
137
+ 2. Combine the SMILES strings, SDF data, and 3D molecular properties for each molecule.
138
+ 3. Split the dataset using random split and scaffold split (train, test, validation)
139
+
140
+ If you would like to try these processes with the original dataset,
141
+ please follow the instructions in the [Preprocessing Script](https://huggingface.co/datasets/maomlab/Molecule3D/blob/main/Molecule3D_preprocessing.py) file located in our Molecule3D repository.
142
+
143
+
144
+
145
+ ## Quickstart Usage
146
+
147
+ ### Load a dataset in python
148
+ Each subset can be loaded into python using the Huggingface [datasets](https://huggingface.co/docs/datasets/index) library.
149
+ First, from the command line install the `datasets` library
150
+
151
+ $ pip install datasets
152
+
153
+ then, from within python load the datasets library
154
+
155
+ >>> import datasets
156
+
157
+ and load one of the `Molecule3D` datasets, e.g.,
158
+
159
+ >>> Molecule3D = datasets.load_dataset('maomlab/Molecule3D', name = 'Molecule3D_random_split') # can put 'Molecule3D_scaffold_split' for the name as well
160
+ README.md: 100% 4.95k/4.95k [00:00<00:00, 559kB/s]
161
+ Generating train split: 100% 2339788/2339788 [00:34<00:00, 85817.85 examples/s]
162
+ Generating test split: 100% 779930/779930 [00:15<00:00, 96660.33 examples/s]
163
+ Generating validation split:  100% 779929/779929 [00:09<00:00, 79064.99 examples/s]
164
+
165
+ and inspecting the dataset
166
+
167
+ >>> Molecule3D
168
+ DatasetDict({
169
+ train: Dataset({
170
+ features: ['index', 'SMILES', 'sdf', 'cid', 'dipole x', 'dipole y', 'dipole z', 'homo', 'lumo', 'Y', 'scf energy'],
171
+ num_rows: 2339788
172
+ })
173
+ test: Dataset({
174
+ features: ['index', 'SMILES', 'sdf', 'cid', 'dipole x', 'dipole y', 'dipole z', 'homo', 'lumo', 'Y', 'scf energy'],
175
+ num_rows: 779930
176
+ })
177
+ validation: Dataset({
178
+ features: ['index', 'SMILES', 'sdf', 'cid', 'dipole x', 'dipole y', 'dipole z', 'homo', 'lumo', 'Y', 'scf energy'],
179
+ num_rows: 779929
180
+ })
181
+ })
182
+
183
+
184
+ ### Use a dataset to train a model
185
+ One way to use the dataset is through the [MolFlux](https://exscientia.github.io/molflux/) package developed by Exscientia.
186
+ First, from the command line, install `MolFlux` library with `catboost` and `rdkit` support
187
+
188
+ pip install 'molflux[catboost,rdkit]'
189
+
190
+ then load, featurize, split, fit, and evaluate the catboost model
191
+
192
+ import json
193
+ from datasets import load_dataset
194
+ from molflux.datasets import featurise_dataset
195
+ from molflux.features import load_from_dicts as load_representations_from_dicts
196
+ from molflux.splits import load_from_dict as load_split_from_dict
197
+ from molflux.modelzoo import load_from_dict as load_model_from_dict
198
+ from molflux.metrics import load_suite
199
+
200
+ split_dataset = load_dataset('maomlab/Molecule3D', name = 'Molecule3D_random_split') # can put 'Molecule3D_scaffold_split' for the name as well
201
+
202
+ split_featurised_dataset = featurise_dataset(
203
+ split_dataset,
204
+ column = "SMILES",
205
+ representations = load_representations_from_dicts([{"name": "morgan"}, {"name": "maccs_rdkit"}]))
206
+
207
+ model = load_model_from_dict({
208
+ "name": "cat_boost_regressor",
209
+ "config": {
210
+ "x_features": ['SMILES::morgan', 'SMILES::maccs_rdkit'],
211
+ "y_features": ['Y']}})
212
+
213
+ model.train(split_featurised_dataset["train"])
214
+ preds = model.predict(split_featurised_dataset["test"])
215
+
216
+ regression_suite = load_suite("regression")
217
+
218
+ scores = regression_suite.compute(
219
+ references=split_featurised_dataset["test"]['Y'],
220
+ predictions=preds["cat_boost_regressor::Y"])
221
+
222
+
223
+ ## Citation
224
+ @misc{https://doi.org/10.48550/arxiv.2110.01717,
225
+ doi = {10.48550/ARXIV.2110.01717},
226
+ url = {https://arxiv.org/abs/2110.01717},
227
+ author = {Xu, Zhao and Luo, Youzhi and Zhang, Xuan and Xu, Xinyi and Xie, Yaochen and Liu, Meng and Dickerson, Kaleb and Deng, Cheng and Nakata, Maho and Ji, Shuiwang},
228
+ keywords = {Machine Learning (cs.LG), Artificial Intelligence (cs.AI), FOS: Computer and information sciences, FOS: Computer and information sciences},
229
+ title = {Molecule3D: A Benchmark for Predicting 3D Geometries from Molecular Graphs},
230
+ publisher = {arXiv},
231
+ year = {2021},
232
+ copyright = {arXiv.org perpetual, non-exclusive license}
233
+ }