├── .gitignore ├── LICENSE.txt ├── README.md ├── __init__.py ├── chemprop ├── __init__.py ├── data │ ├── __init__.py │ ├── data.py │ ├── scaffold.py │ ├── scaler.py │ ├── similarity.py │ ├── unsupervised_cluster.py │ ├── utils.py │ └── vocab.py ├── data_processing │ ├── __init__.py │ ├── avg_dups.py │ ├── plot_distribution.py │ ├── process_zinc.py │ └── resplit.py ├── features │ ├── __init__.py │ ├── async_featurization.py │ ├── descriptors.py │ ├── featurization.py │ ├── functional_groups.py │ ├── kernels.py │ ├── morgan_fingerprint.py │ ├── rdkit_features.py │ ├── rdkit_normalized_features.py │ ├── smarts.txt │ └── utils.py ├── models │ ├── __init__.py │ ├── gan.py │ ├── jtnn.py │ ├── learned_kernel.py │ ├── model.py │ ├── moe.py │ └── mpn.py ├── nn_utils.py ├── parsing.py ├── random_forest.py ├── train │ ├── __init__.py │ ├── cross_validate.py │ ├── evaluate.py │ ├── make_predictions.py │ ├── predict.py │ ├── run_training.py │ └── train.py └── utils.py ├── data ├── all_smiles.csv.gz ├── bace.csv ├── bbbp.csv ├── clintox.csv ├── delaney.csv ├── freesolv.csv ├── hiv.csv.gz ├── lipo.csv ├── muv.csv.gz ├── pcba.csv.gz ├── pdbbind_core.csv ├── pdbbind_full.csv ├── pdbbind_refined.csv ├── qm7.csv.gz ├── qm8.csv.gz ├── qm9.csv.gz ├── sanitize.py ├── sider.csv ├── tox21.csv └── toxcast.csv.gz ├── distributions ├── chembl │ └── chembl_labels_log.png ├── delaney │ └── delaney_logp.png ├── freesolv │ └── freesolv_freesolv.png ├── lipo │ └── lipo_lipo.png ├── qm8 │ ├── qm8_E1-CAM.png │ ├── qm8_E1-CC2.png │ ├── qm8_E1-PBE0.png │ ├── qm8_E2-CAM.png │ ├── qm8_E2-CC2.png │ ├── qm8_E2-PBE0.png │ ├── qm8_f1-CAM.png │ ├── qm8_f1-CC2.png │ ├── qm8_f1-PBE0.png │ ├── qm8_f2-CAM.png │ ├── qm8_f2-CC2.png │ └── qm8_f2-PBE0.png └── qm9 │ ├── qm9_alpha.png │ ├── qm9_cv.png │ ├── qm9_g298.png │ ├── qm9_g298_atom.png │ ├── qm9_gap.png │ ├── qm9_h298.png │ ├── qm9_h298_atom.png │ ├── qm9_homo.png │ ├── qm9_lumo.png │ ├── qm9_mu.png │ ├── qm9_r2.png │ ├── qm9_u0.png │ ├── qm9_u0_atom.png │ ├── qm9_u298.png │ ├── qm9_u298_atom.png │ └── qm9_zpve.png ├── hyperparameter_optimization.py ├── model_comparison.py ├── predict.py ├── random_forest.py ├── requirements.txt ├── scripts ├── __init__.py ├── avg_dups.py ├── filter_by_scaffold.py ├── overlap.py ├── plot_distribution.py ├── resplit_data.py ├── save_features.py ├── similarity.py ├── visualize_encoding_property_space.py ├── viz_attention.py └── vocab.py ├── setup.cfg ├── setup.py ├── static ├── jsme │ ├── .directory │ ├── JME_to_JSME.html │ ├── JME_to_JSME_simple.html │ ├── JSME.html │ ├── JSME_SVG.html │ ├── JSME_atom_highlight_demo.html │ ├── JSME_autoresize.html │ ├── JSME_callback_and_star.html │ ├── JSME_chemical_resolver_demo.html │ ├── JSME_depict.html │ ├── JSME_depict_action.html │ ├── JSME_depict_action_callback.html │ ├── JSME_depict_edit_toggle.html │ ├── JSME_depict_smiles.html │ ├── JSME_dnd_demo.html │ ├── JSME_editor_plus_SVG.html │ ├── JSME_hidden.html │ ├── JSME_minimal.html │ ├── JSME_minimal2.html │ ├── JSME_minimal3.html │ ├── JSME_parent_and_metabolites.html │ ├── JSME_resize.html │ ├── JSME_smiles_atom_highlight.html │ ├── JSME_template.html │ ├── JSME_test.html │ ├── api_javadoc │ │ ├── allclasses-frame.html │ │ ├── allclasses-noframe.html │ │ ├── constant-values.html │ │ ├── deprecated-list.html │ │ ├── export │ │ │ └── client │ │ │ │ ├── JSME.html │ │ │ │ ├── Utils.html │ │ │ │ ├── package-frame.html │ │ │ │ ├── package-summary.html │ │ │ │ └── package-tree.html │ │ ├── help-doc.html │ │ ├── index-all.html │ │ ├── index.html │ │ ├── overview-tree.html │ │ ├── package-list │ │ ├── script.js │ │ └── stylesheet.css │ ├── bootstrap.css │ ├── bootstrap.html │ ├── doc.css │ ├── doc.html │ ├── index.html │ ├── jme_examples │ │ ├── jme_example1.html │ │ ├── jme_example2.html │ │ ├── jme_example3.html │ │ ├── jme_example4.html │ │ └── jme_window.html │ ├── jsme │ │ ├── 222ADBFEC322C2723C6ED2C4FB31B217.cache.js │ │ ├── 293DFEFA807A962F28C09E358B34A434.cache.js │ │ ├── 396F806CD63ABD414BFBB9D57429F05B.cache.png │ │ ├── 40BAF81124143A595056A9CCA0E9DBBA.cache.png │ │ ├── 4841BDE9DC293BA35F7762B4D8EFD236.cache.png │ │ ├── 61B683D3493CAED438D5743A0404863D.cache.js │ │ ├── 6ABB8447ACAB1353A478923AC9C0550B.cache.js │ │ ├── 7A65B607B90DE29D7EA26AA83BF69D4F.cache.js │ │ ├── 8816D61E367E34DBCFA53666849E21D8.cache.js │ │ ├── 8BDB7ED57B756F8D50277056A0D59DA8.cache.js │ │ ├── 96E40B969193BD74B8A621486920E79C.cache.js │ │ ├── A2384E54F71557BAEA414A43D47F17EA.cache.js │ │ ├── A6DBDE07E3A8F66E8959A4F32505E16B.cache.png │ │ ├── C8A71BD2E1367E9BB43A1B9C25871BEE.cache.js │ │ ├── C9EEF554958AACEE6A060F620375E4FA.cache.js │ │ ├── D4DF9EC9DD21B943E35F3D5696D5D2A1.cache.js │ │ ├── D9A64F1634E29088B910B3E0D4621E49.cache.js │ │ ├── DF7764EEC1903CD03C9545B354D8D8E4.cache.png │ │ ├── clear.cache.gif │ │ ├── compilation-mappings.txt │ │ ├── deferredjs │ │ │ ├── 222ADBFEC322C2723C6ED2C4FB31B217 │ │ │ │ ├── 1.cache.js │ │ │ │ ├── 2.cache.js │ │ │ │ ├── 3.cache.js │ │ │ │ ├── 4.cache.js │ │ │ │ ├── 5.cache.js │ │ │ │ ├── 6.cache.js │ │ │ │ ├── 7.cache.js │ │ │ │ ├── 8.cache.js │ │ │ │ └── 9.cache.js │ │ │ ├── 293DFEFA807A962F28C09E358B34A434 │ │ │ │ ├── 1.cache.js │ │ │ │ ├── 2.cache.js │ │ │ │ ├── 3.cache.js │ │ │ │ ├── 4.cache.js │ │ │ │ ├── 5.cache.js │ │ │ │ ├── 6.cache.js │ │ │ │ ├── 7.cache.js │ │ │ │ ├── 8.cache.js │ │ │ │ └── 9.cache.js │ │ │ ├── 61B683D3493CAED438D5743A0404863D │ │ │ │ ├── 1.cache.js │ │ │ │ ├── 2.cache.js │ │ │ │ ├── 3.cache.js │ │ │ │ ├── 4.cache.js │ │ │ │ ├── 5.cache.js │ │ │ │ ├── 6.cache.js │ │ │ │ ├── 7.cache.js │ │ │ │ ├── 8.cache.js │ │ │ │ └── 9.cache.js │ │ │ ├── 6ABB8447ACAB1353A478923AC9C0550B │ │ │ │ ├── 1.cache.js │ │ │ │ ├── 2.cache.js │ │ │ │ ├── 3.cache.js │ │ │ │ ├── 4.cache.js │ │ │ │ ├── 5.cache.js │ │ │ │ ├── 6.cache.js │ │ │ │ ├── 7.cache.js │ │ │ │ ├── 8.cache.js │ │ │ │ └── 9.cache.js │ │ │ ├── 7A65B607B90DE29D7EA26AA83BF69D4F │ │ │ │ ├── 1.cache.js │ │ │ │ ├── 2.cache.js │ │ │ │ ├── 3.cache.js │ │ │ │ ├── 4.cache.js │ │ │ │ ├── 5.cache.js │ │ │ │ ├── 6.cache.js │ │ │ │ ├── 7.cache.js │ │ │ │ ├── 8.cache.js │ │ │ │ └── 9.cache.js │ │ │ ├── 8816D61E367E34DBCFA53666849E21D8 │ │ │ │ ├── 1.cache.js │ │ │ │ ├── 2.cache.js │ │ │ │ ├── 3.cache.js │ │ │ │ ├── 4.cache.js │ │ │ │ ├── 5.cache.js │ │ │ │ ├── 6.cache.js │ │ │ │ ├── 7.cache.js │ │ │ │ ├── 8.cache.js │ │ │ │ └── 9.cache.js │ │ │ ├── 8BDB7ED57B756F8D50277056A0D59DA8 │ │ │ │ ├── 1.cache.js │ │ │ │ ├── 2.cache.js │ │ │ │ ├── 3.cache.js │ │ │ │ ├── 4.cache.js │ │ │ │ ├── 5.cache.js │ │ │ │ ├── 6.cache.js │ │ │ │ ├── 7.cache.js │ │ │ │ ├── 8.cache.js │ │ │ │ └── 9.cache.js │ │ │ ├── A2384E54F71557BAEA414A43D47F17EA │ │ │ │ ├── 1.cache.js │ │ │ │ ├── 2.cache.js │ │ │ │ ├── 3.cache.js │ │ │ │ ├── 4.cache.js │ │ │ │ ├── 5.cache.js │ │ │ │ ├── 6.cache.js │ │ │ │ ├── 7.cache.js │ │ │ │ ├── 8.cache.js │ │ │ │ └── 9.cache.js │ │ │ ├── C8A71BD2E1367E9BB43A1B9C25871BEE │ │ │ │ ├── 1.cache.js │ │ │ │ ├── 2.cache.js │ │ │ │ ├── 3.cache.js │ │ │ │ ├── 4.cache.js │ │ │ │ ├── 5.cache.js │ │ │ │ ├── 6.cache.js │ │ │ │ ├── 7.cache.js │ │ │ │ ├── 8.cache.js │ │ │ │ └── 9.cache.js │ │ │ ├── C9EEF554958AACEE6A060F620375E4FA │ │ │ │ ├── 1.cache.js │ │ │ │ ├── 2.cache.js │ │ │ │ ├── 3.cache.js │ │ │ │ ├── 4.cache.js │ │ │ │ ├── 5.cache.js │ │ │ │ ├── 6.cache.js │ │ │ │ ├── 7.cache.js │ │ │ │ ├── 8.cache.js │ │ │ │ └── 9.cache.js │ │ │ ├── D4DF9EC9DD21B943E35F3D5696D5D2A1 │ │ │ │ ├── 1.cache.js │ │ │ │ ├── 2.cache.js │ │ │ │ ├── 3.cache.js │ │ │ │ ├── 4.cache.js │ │ │ │ ├── 5.cache.js │ │ │ │ ├── 6.cache.js │ │ │ │ ├── 7.cache.js │ │ │ │ ├── 8.cache.js │ │ │ │ └── 9.cache.js │ │ │ └── D9A64F1634E29088B910B3E0D4621E49 │ │ │ │ ├── 1.cache.js │ │ │ │ ├── 2.cache.js │ │ │ │ ├── 3.cache.js │ │ │ │ ├── 4.cache.js │ │ │ │ ├── 5.cache.js │ │ │ │ ├── 6.cache.js │ │ │ │ ├── 7.cache.js │ │ │ │ ├── 8.cache.js │ │ │ │ └── 9.cache.js │ │ ├── gwt │ │ │ └── chrome │ │ │ │ ├── chrome.css │ │ │ │ ├── chrome_rtl.css │ │ │ │ ├── images │ │ │ │ ├── button │ │ │ │ │ ├── menu-button-arrow-disabled.png │ │ │ │ │ ├── menu-button-arrow.png │ │ │ │ │ ├── split-button-arrow-active.png │ │ │ │ │ ├── split-button-arrow-disabled.png │ │ │ │ │ ├── split-button-arrow-focus.png │ │ │ │ │ ├── split-button-arrow-hover.png │ │ │ │ │ └── split-button-arrow.png │ │ │ │ ├── combobox │ │ │ │ │ ├── arrow-down-disabled.png │ │ │ │ │ ├── arrow-down.png │ │ │ │ │ ├── ellipsis-disabled.png │ │ │ │ │ └── ellipsis.png │ │ │ │ ├── corner.png │ │ │ │ ├── corner_ie6.png │ │ │ │ ├── fastree │ │ │ │ │ ├── selectionBar.gif │ │ │ │ │ ├── treeClosed.gif │ │ │ │ │ ├── treeLoading.gif │ │ │ │ │ └── treeOpen.gif │ │ │ │ ├── glasspanel │ │ │ │ │ └── blue_ridge.png │ │ │ │ ├── hborder.png │ │ │ │ ├── hborder_ie6.png │ │ │ │ ├── ie6 │ │ │ │ │ ├── corner_dialog_topleft.png │ │ │ │ │ ├── corner_dialog_topright.png │ │ │ │ │ ├── hborder_blue_shadow.png │ │ │ │ │ ├── hborder_gray_shadow.png │ │ │ │ │ ├── vborder_blue_shadow.png │ │ │ │ │ └── vborder_gray_shadow.png │ │ │ │ ├── scrolltable │ │ │ │ │ └── bg_header_gradient.gif │ │ │ │ ├── splitPanelThumb.png │ │ │ │ ├── valuespinner │ │ │ │ │ └── bg_textbox.png │ │ │ │ ├── vborder.png │ │ │ │ └── vborder_ie6.png │ │ │ │ ├── mosaic.css │ │ │ │ └── mosaic_rtl.css │ │ ├── jsa.css │ │ ├── jsme.devmode.js │ │ └── jsme.nocache.js │ ├── license.txt │ ├── release_notes.txt │ └── test_depict_many_smiles_in_table.html ├── message_passing.png └── stylesheets │ └── style.css ├── templates ├── checkpoints.html ├── data.html ├── home.html ├── layout.html ├── macros.html ├── predict.html └── train.html ├── train.py └── web.py /.gitignore: -------------------------------------------------------------------------------- 1 | *.pyc 2 | deepchem-test/jtnn 3 | deepchem-test/data 4 | *.idea 5 | *.DS_Store 6 | *.vscode 7 | -------------------------------------------------------------------------------- /LICENSE.txt: -------------------------------------------------------------------------------- 1 | MIT License 2 | 3 | Copyright (c) 2018 Wengong Jin, Kyle Swanson, Kevin Yang, Regina Barzilay, Tommi Jaakkola 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. 22 | -------------------------------------------------------------------------------- /__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/__init__.py -------------------------------------------------------------------------------- /chemprop/__init__.py: -------------------------------------------------------------------------------- 1 | import chemprop.data 2 | import chemprop.data_processing 3 | import chemprop.features 4 | import chemprop.models 5 | import chemprop.train 6 | 7 | import chemprop.nn_utils 8 | import chemprop.parsing 9 | import chemprop.utils 10 | -------------------------------------------------------------------------------- /chemprop/data/__init__.py: -------------------------------------------------------------------------------- 1 | from .data import MoleculeDatapoint, MoleculeDataset 2 | from .scaffold import cluster_split 3 | from .scaler import StandardScaler 4 | from .similarity import morgan_similarity, scaffold_similarity 5 | from .unsupervised_cluster import generate_unsupervised_cluster_labels 6 | from .vocab import get_vocab, load_vocab 7 | -------------------------------------------------------------------------------- /chemprop/data/scaler.py: -------------------------------------------------------------------------------- 1 | from typing import List 2 | 3 | import numpy as np 4 | 5 | 6 | class StandardScaler: 7 | def __init__(self, means: np.ndarray = None, stds: np.ndarray = None, replace_nan_token=None): 8 | """Initialize StandardScaler, optionally with means and standard deviations precomputed.""" 9 | self.means = means 10 | self.stds = stds 11 | self.replace_nan_token = replace_nan_token 12 | 13 | def fit(self, X: List[List[float]]) -> 'StandardScaler': 14 | """ 15 | Learns means and standard deviations across the 0-th axis. 16 | 17 | :param X: A list of lists of floats. 18 | :return: The fitted StandardScaler. 19 | """ 20 | X = np.array(X).astype(float) 21 | self.means = np.nanmean(X, axis=0) 22 | self.stds = np.nanstd(X, axis=0) 23 | self.means = np.where(np.isnan(self.means), np.zeros(self.means.shape), self.means) 24 | self.stds = np.where(np.isnan(self.stds), np.ones(self.stds.shape), self.stds) 25 | self.stds = np.where(self.stds == 0, np.ones(self.stds.shape), self.stds) 26 | 27 | return self 28 | 29 | def transform(self, X: List[List[float]]): 30 | """ 31 | Transforms the data by subtracting the means and dividing by the standard deviations. 32 | 33 | :param X: A list of lists of floats. 34 | :return: The transformed data. 35 | """ 36 | X = np.array(X).astype(float) 37 | transformed_with_nan = (X - self.means) / self.stds 38 | transformed_with_none = np.where(np.isnan(transformed_with_nan), self.replace_nan_token, transformed_with_nan) 39 | 40 | return transformed_with_none 41 | 42 | def inverse_transform(self, X: List[List[float]]): 43 | """ 44 | Performs the inverse transformation by multiplying by the standard deviations and adding the means. 45 | 46 | :param X: A list of lists of floats. 47 | :return: The inverse transformed data. 48 | """ 49 | X = np.array(X).astype(float) 50 | transformed_with_nan = X * self.stds + self.means 51 | transformed_with_none = np.where(np.isnan(transformed_with_nan), self.replace_nan_token, transformed_with_nan) 52 | 53 | return transformed_with_none 54 | -------------------------------------------------------------------------------- /chemprop/data/unsupervised_cluster.py: -------------------------------------------------------------------------------- 1 | from logging import Logger 2 | 3 | from sklearn.cluster import MiniBatchKMeans 4 | import torch 5 | from tqdm import trange 6 | 7 | from .data import MoleculeDataset 8 | 9 | 10 | def get_cluster_labels(encodings, n_clusters: int = 10000, seed: int = 0, logger: Logger = None): 11 | n_clusters = int(min(n_clusters, len(encodings)/10)) # so we don't crash if we only picked a small number of encodings 12 | kmeans = MiniBatchKMeans(n_clusters=n_clusters, random_state=seed) 13 | cluster_labels = kmeans.fit_predict(encodings) 14 | return cluster_labels 15 | 16 | 17 | def generate_unsupervised_cluster_labels(model, data, args, logger: Logger = None): 18 | encodings = [] 19 | for i in trange(0, len(data), args.batch_size): 20 | batch = MoleculeDataset(data[i:i + args.batch_size]) 21 | with torch.no_grad(): 22 | encodings.extend([enc for enc in model.encoder(batch.smiles()).cpu().numpy()]) 23 | cluster_labels = get_cluster_labels(encodings, n_clusters=args.unsupervised_n_clusters, logger=logger) 24 | cluster_labels = cluster_labels.reshape(-1, 1).astype(int).tolist() 25 | data.set_targets(cluster_labels) 26 | 27 | -------------------------------------------------------------------------------- /chemprop/data_processing/__init__.py: -------------------------------------------------------------------------------- 1 | from .avg_dups import average_duplicates 2 | from .plot_distribution import plot_distribution 3 | from .resplit import resplit -------------------------------------------------------------------------------- /chemprop/data_processing/avg_dups.py: -------------------------------------------------------------------------------- 1 | """Averages duplicate data points in a dataset.""" 2 | 3 | from collections import defaultdict 4 | 5 | import numpy as np 6 | 7 | from chemprop.data.utils import get_data, get_header 8 | 9 | 10 | def average_duplicates(args): 11 | print('Loading data') 12 | header = get_header(args.data_path) 13 | data = get_data(path=args.data_path) 14 | print(f'Data size = {len(data):,}') 15 | 16 | # Map SMILES string to lists of targets 17 | smiles_in_order = [] 18 | smiles_to_targets = defaultdict(list) 19 | for smiles, targets in zip(data.smiles(), data.targets()): 20 | smiles_to_targets[smiles].append(targets) 21 | if len(smiles_to_targets[smiles]) == 1: 22 | smiles_in_order.append(smiles) 23 | 24 | # Find duplicates 25 | duplicate_count = 0 26 | stds = [] 27 | new_data = [] 28 | for smiles in smiles_in_order: 29 | all_targets = smiles_to_targets[smiles] 30 | duplicate_count += len(all_targets) - 1 31 | num_tasks = len(all_targets[0]) 32 | 33 | targets_by_task = [[] for _ in range(num_tasks)] 34 | for task in range(num_tasks): 35 | for targets in all_targets: 36 | if targets[task] is not None: 37 | targets_by_task[task].append(targets[task]) 38 | 39 | stds.append([np.std(task_targets) if len(task_targets) > 0 else 0.0 for task_targets in targets_by_task]) 40 | means = [np.mean(task_targets) if len(task_targets) > 0 else None for task_targets in targets_by_task] 41 | new_data.append((smiles, means)) 42 | 43 | print(f'Number of duplicates = {duplicate_count:,}') 44 | print(f'Duplicate standard deviation per task = {", ".join(f":{std:.4e}" for std in np.mean(stds, axis=0))}') 45 | print(f'New data size = {len(new_data):,}') 46 | 47 | # Save new data 48 | with open(args.save_path, 'w') as f: 49 | f.write(','.join(header) + '\n') 50 | 51 | for smiles, avg_targets in new_data: 52 | f.write(smiles + ',' + ','.join(str(value) if value is not None else '' for value in avg_targets) + '\n') 53 | -------------------------------------------------------------------------------- /chemprop/data_processing/plot_distribution.py: -------------------------------------------------------------------------------- 1 | import os 2 | 3 | import matplotlib.pyplot as plt 4 | 5 | from chemprop.data.utils import get_data, get_task_names 6 | 7 | 8 | def plot_distribution(data_path: str, save_dir: str, bins: int): 9 | """ 10 | Plots the distribution of values of a dataset. 11 | 12 | :param data_path: Path to data CSV file. 13 | :param save_dir: Directory where plot PNGs will be saved. 14 | :param bins: Number of bins in histogram. 15 | """ 16 | # Get values 17 | task_names = get_task_names(data_path) 18 | data = get_data(path=data_path) 19 | targets = data.targets() 20 | 21 | # Arrange values by task 22 | data_size, num_tasks = len(targets), len(task_names) 23 | values = [[targets[i][j] for i in range(data_size)] for j in range(num_tasks)] 24 | 25 | # Plot distributions for each task 26 | data_name = os.path.basename(data_path).replace('.csv', '') 27 | 28 | for i in range(num_tasks): 29 | plt.clf() 30 | plt.hist(values[i], bins=bins) 31 | 32 | # Save plot 33 | plt.title(f'{data_name} - {task_names[i]}') 34 | plt.xlabel(task_names[i]) 35 | plt.ylabel('Frequency') 36 | plt.savefig(os.path.join(save_dir, f'{data_name}_{task_names[i]}.png')) 37 | -------------------------------------------------------------------------------- /chemprop/data_processing/process_zinc.py: -------------------------------------------------------------------------------- 1 | import os 2 | from tqdm import tqdm 3 | from argparse import ArgumentParser 4 | 5 | if __name__ == '__main__': 6 | parser = ArgumentParser() 7 | parser.add_argument('--zinc_dir', type=str, required=True, 8 | help='Path to dir with raw ZINC files') 9 | parser.add_argument('--max_size', type=int, default=0, 10 | help='Max number of smiles (0 = all)') 11 | parser.add_argument('--write_loc', type=str, required=True, 12 | help='Where to write to') 13 | parser.add_argument('--individual_files', action='store_true', default=False, 14 | help='Process files individually') 15 | args = parser.parse_args() 16 | 17 | if args.individual_files: 18 | os.makedirs(args.write_loc, exist_ok=True) 19 | for root, _, names in os.walk(args.zinc_dir): 20 | for name in tqdm(names, total=len(names)): 21 | _, ext = os.path.splitext(name) 22 | if ext == '.txt': 23 | with open(os.path.join(root, name), 'r') as rf, \ 24 | open(os.path.join(args.write_loc, name), 'w') as wf: 25 | wf.write('smiles') 26 | wf.write('\n') 27 | rf.readline() 28 | for line in rf: 29 | smiles = line.strip().split('\t')[0] 30 | wf.write(smiles) 31 | wf.write('\n') 32 | 33 | else: 34 | with open(os.path.join(args.write_loc), 'w') as wf: 35 | wf.write('smiles') 36 | wf.write('\n') 37 | count = 0 38 | for root, _, names in os.walk(args.zinc_dir): 39 | for name in tqdm(names, total=len(names)): 40 | _, ext = os.path.splitext(name) 41 | if ext == '.txt': 42 | with open(os.path.join(root, name), 'r') as rf: 43 | rf.readline() 44 | for line in rf: 45 | if count > args.max_size: 46 | break 47 | smiles = line.strip().split('\t')[0] 48 | wf.write(smiles) 49 | wf.write('\n') 50 | count += 1 -------------------------------------------------------------------------------- /chemprop/data_processing/resplit.py: -------------------------------------------------------------------------------- 1 | from argparse import Namespace 2 | 3 | 4 | def resplit(args: Namespace): 5 | """ 6 | Resplits the train and validation data chronologically. 7 | 8 | Assumes that the data at train_path and val_path are sorted 9 | chronologically within each file but have been split randomly 10 | between the two files. This function puts the first (1 - val_frac) 11 | of both the train and validation data in the new train_save file 12 | and puts the remaining val_frac of both in the new val_save file. 13 | That way, the new validation data comes chronologically after 14 | the new training data. 15 | """ 16 | train_frac = 1 - args.val_frac 17 | 18 | # Get train and validation sizes 19 | with open(args.train_path, 'r') as f: 20 | train_len = sum(1 for _ in f) - 1 21 | with open(args.val_path, 'r') as f: 22 | val_len = sum(1 for _ in f) - 1 23 | 24 | # Resplit data 25 | with open(args.train_path, 'r') as rtf, open(args.val_path, 'r') as rvf, \ 26 | open(args.train_save, 'w') as wtf, open(args.val_save, 'w') as wvf: 27 | header = rtf.readline().strip() 28 | rvf.readline() # skip header 29 | 30 | wtf.write(header + '\n') 31 | wvf.write(header + '\n') 32 | 33 | for i in range(train_len): 34 | line = rtf.readline().strip() 35 | 36 | if i < train_frac * train_len: 37 | wtf.write(line + '\n') 38 | else: 39 | wvf.write(line + '\n') 40 | 41 | for i in range(val_len): 42 | line = rvf.readline().strip() 43 | 44 | if i < train_frac * val_len: 45 | wtf.write(line + '\n') 46 | else: 47 | wvf.write(line + '\n') 48 | -------------------------------------------------------------------------------- /chemprop/features/__init__.py: -------------------------------------------------------------------------------- 1 | from .featurization import atom_features, bond_features, BatchMolGraph, get_atom_fdim, get_bond_fdim, mol2graph 2 | from .functional_groups import FunctionalGroupFeaturizer 3 | from .kernels import get_kernel_func 4 | from .morgan_fingerprint import morgan_fingerprint 5 | from .rdkit_features import rdkit_2d_features 6 | from .utils import load_features, get_features_func 7 | -------------------------------------------------------------------------------- /chemprop/features/async_featurization.py: -------------------------------------------------------------------------------- 1 | from argparse import Namespace 2 | from typing import List, Tuple, Union 3 | 4 | import numpy as np 5 | from rdkit import Chem 6 | from rdkit.Chem import AllChem 7 | 8 | import torch 9 | torch.multiprocessing.set_sharing_strategy('file_system') 10 | from torch.multiprocessing import Queue, Pool 11 | 12 | from chemprop.data import MoleculeDataset 13 | from .featurization import mol2graph 14 | 15 | def mol2graph_helper(pair): 16 | batch, args = pair 17 | return batch, mol2graph(batch.smiles(), args) 18 | 19 | def async_mol2graph(q: Queue, 20 | data: MoleculeDataset, 21 | args: Namespace, 22 | num_iters: int, 23 | iter_size: int, 24 | exit_q: Queue, 25 | last_batch: bool=False): 26 | batches = [] 27 | for i in range(0, num_iters, iter_size): # will only go up to max size of queue, then yield 28 | if not last_batch and i + args.batch_size > len(data): 29 | break 30 | batch = MoleculeDataset(data[i:i + args.batch_size]) 31 | batches.append(batch) 32 | if len(batches) == args.batches_per_queue_group: # many at a time, since synchronization is expensive 33 | with Pool() as pool: 34 | processed_batches = pool.map(mol2graph_helper, [(batch, args) for batch in batches]) 35 | q.put(processed_batches) 36 | batches = [] 37 | if len(batches) > 0: 38 | with Pool() as pool: 39 | processed_batches = pool.map(mol2graph_helper, [(batch, args) for batch in batches]) 40 | q.put(processed_batches) 41 | exit_q.get() # prevent from exiting until main process tells it to; otherwise we apparently can't read the end of the queue and crash -------------------------------------------------------------------------------- /chemprop/features/descriptors.py: -------------------------------------------------------------------------------- 1 | from mordred import Calculator, descriptors 2 | import numpy as np 3 | from rdkit import Chem 4 | 5 | 6 | mordred_calc = Calculator(descriptors, ignore_3D=True) # can't do 3D without sdf or mol file 7 | 8 | 9 | def mordred_features(mol: Chem.Mol) -> np.ndarray: 10 | return np.array([float(f) for f in mordred_calc(mol)]) 11 | -------------------------------------------------------------------------------- /chemprop/features/functional_groups.py: -------------------------------------------------------------------------------- 1 | from argparse import Namespace 2 | import numpy as np 3 | from typing import List, Union 4 | 5 | from rdkit import Chem 6 | 7 | 8 | def get_num_functional_groups(args: Namespace): 9 | with open(args.functional_group_smarts, 'r') as f: 10 | count = len(f.readlines()) 11 | return count 12 | 13 | 14 | class FunctionalGroupFeaturizer: 15 | """ 16 | Class for extracting feature vector of indicators for atoms being parts of functional groups. 17 | """ 18 | def __init__(self, args: Namespace): 19 | self.smarts = [] 20 | with open(args.functional_group_smarts, 'r') as f: 21 | for line in f: 22 | self.smarts.append(Chem.MolFromSmarts(line.strip())) 23 | 24 | def featurize(self, smiles: Union[Chem.Mol, str]) -> List[List[int]]: 25 | """ 26 | Given a molecule in SMILES form, return a feature vector of indicators for each atom, 27 | indicating whether the atom is part of each functional group. 28 | Can also directly accept a Chem molecule. 29 | Searches through the functional groups given in smarts.txt. 30 | 31 | :param smiles: A smiles string representing a molecule. 32 | :return: Numpy array of shape num_atoms x num_features (num functional groups) 33 | """ 34 | if type(smiles) == str: 35 | mol = Chem.MolFromSmiles(smiles) # turns out rdkit knows to match even without adding Hs 36 | else: 37 | mol = smiles 38 | features = np.zeros((mol.GetNumAtoms(), len(self.smarts))) # num atoms (without Hs) x num features 39 | for i, smarts in enumerate(self.smarts): 40 | for group in mol.GetSubstructMatches(smarts): 41 | for idx in group: 42 | features[idx][i] = 1 43 | 44 | return features.tolist() 45 | 46 | 47 | if __name__ == '__main__': 48 | featurizer = FunctionalGroupFeaturizer() 49 | features = featurizer.featurize('C(#N)C(=O)C#N') 50 | -------------------------------------------------------------------------------- /chemprop/features/morgan_fingerprint.py: -------------------------------------------------------------------------------- 1 | import numpy as np 2 | from rdkit import Chem, DataStructs 3 | from rdkit.Chem import AllChem 4 | 5 | 6 | def morgan_fingerprint(smiles: str, radius: int = 2, num_bits: int = 2048, use_counts: bool = False) -> np.ndarray: 7 | """ 8 | Generates a morgan fingerprint for a smiles string. 9 | 10 | :param smiles: A smiles string for a molecule. 11 | :param radius: The radius of the fingerprint. 12 | :param num_bits: The number of bits to use in the fingerprint. 13 | :param use_counts: Whether to use counts or just a bit vector for the fingerprint 14 | :return: A 1-D numpy array containing the morgan fingerprint. 15 | """ 16 | if type(smiles) == str: 17 | mol = Chem.MolFromSmiles(smiles) 18 | else: 19 | mol = smiles 20 | if use_counts: 21 | fp_vect = AllChem.GetHashedMorganFingerprint(mol, radius, nBits=num_bits) 22 | else: 23 | fp_vect = AllChem.GetMorganFingerprintAsBitVect(mol, radius, nBits=num_bits) 24 | fp = np.zeros((1,)) 25 | DataStructs.ConvertToNumpyArray(fp_vect, fp) 26 | 27 | return fp 28 | -------------------------------------------------------------------------------- /chemprop/features/rdkit_features.py: -------------------------------------------------------------------------------- 1 | from argparse import Namespace 2 | import numpy as np 3 | 4 | from rdkit import Chem 5 | # from rdkit.Chem.rdPartialCharges import ComputeGasteigerCharges 6 | from rdkit.Chem.GraphDescriptors import BalabanJ, BertzCT, Ipc, Chi0, Chi0n, Chi0v, Chi1, Chi1n, Chi1v, Chi2n, Chi2v, \ 7 | Chi3n, Chi3v, Chi4n, Chi4v, HallKierAlpha, Kappa1, Kappa2, Kappa3 8 | from rdkit.Chem.Crippen import MolLogP, MolMR 9 | from rdkit.Chem.Descriptors import MolWt, ExactMolWt, HeavyAtomMolWt, NumValenceElectrons 10 | from rdkit.Chem.Lipinski import HeavyAtomCount, NHOHCount, NOCount, NumHAcceptors, NumHDonors, NumHeteroatoms, \ 11 | NumRotatableBonds, NumAromaticRings, NumSaturatedRings, NumAliphaticRings, NumAromaticHeterocycles, \ 12 | NumAromaticCarbocycles, NumSaturatedHeterocycles, NumSaturatedCarbocycles, NumAliphaticHeterocycles, \ 13 | NumAliphaticCarbocycles, RingCount, FractionCSP3 14 | from rdkit.Chem.rdMolDescriptors import CalcNumSpiroAtoms, CalcNumBridgeheadAtoms, CalcTPSA, CalcLabuteASA, PEOE_VSA_, \ 15 | SMR_VSA_, SlogP_VSA_, MQNs_, CalcAUTOCORR2D, CalcNumAmideBonds 16 | from rdkit.Chem.Fragments import fr_ketone_Topliss 17 | from rdkit.Chem.EState.EState_VSA import EState_VSA1, EState_VSA2, EState_VSA3, EState_VSA4, EState_VSA5, EState_VSA6, \ 18 | EState_VSA7, EState_VSA8, EState_VSA9, EState_VSA10, EState_VSA11, VSA_EState1, VSA_EState2, VSA_EState3, \ 19 | VSA_EState4, VSA_EState5, VSA_EState6, VSA_EState7, VSA_EState8, VSA_EState9, VSA_EState10 20 | 21 | from chemprop.features.functional_groups import FunctionalGroupFeaturizer 22 | 23 | FEATURE_FUNCTIONS = [ 24 | # ComputeGasteigerCharges, 25 | BalabanJ, BertzCT, Ipc, Chi0, Chi0n, Chi0v, Chi1, Chi1n, Chi1v, Chi2n, Chi2v, Chi3n, Chi3v, Chi4n, Chi4v, 26 | HallKierAlpha, Kappa1, Kappa2, Kappa3, 27 | MolLogP, MolMR, 28 | MolWt, ExactMolWt, HeavyAtomMolWt, NumValenceElectrons, 29 | HeavyAtomCount, NHOHCount, NOCount, NumHAcceptors, NumHDonors, NumHeteroatoms, NumRotatableBonds, NumAromaticRings, 30 | NumSaturatedRings, NumAliphaticRings, NumAromaticHeterocycles, NumAromaticCarbocycles, NumSaturatedHeterocycles, 31 | NumSaturatedCarbocycles, NumAliphaticHeterocycles, NumAliphaticCarbocycles, RingCount, FractionCSP3, 32 | CalcNumSpiroAtoms, CalcNumBridgeheadAtoms, CalcTPSA, CalcLabuteASA, PEOE_VSA_, SMR_VSA_, SlogP_VSA_, MQNs_, 33 | CalcAUTOCORR2D, CalcNumAmideBonds, 34 | fr_ketone_Topliss, 35 | EState_VSA1, EState_VSA2, EState_VSA3, EState_VSA4, EState_VSA5, EState_VSA6, EState_VSA7, EState_VSA8, EState_VSA9, 36 | EState_VSA10, EState_VSA11, VSA_EState1, VSA_EState2, VSA_EState3, VSA_EState4, VSA_EState5, VSA_EState6, 37 | VSA_EState7, VSA_EState8, VSA_EState9, VSA_EState10 38 | ] 39 | 40 | 41 | def rdkit_2d_features(smiles: str, args: Namespace): 42 | if type(smiles) == str: 43 | mol = Chem.MolFromSmiles(smiles) 44 | else: 45 | mol = smiles 46 | features = [] 47 | for f in FEATURE_FUNCTIONS: 48 | try: 49 | feature = f(mol) 50 | except: # very very rarely, something like BalabanJ crashes 51 | dummy_mol = Chem.MolFromSmiles('c1ccc2cc(CC3=NCCN3)ccc2c1') 52 | feature = f(dummy_mol) 53 | if type(feature) == list: 54 | features.extend(feature) 55 | else: 56 | features.append(feature) 57 | 58 | # TODO: these following ones take about half of the computation time, can we make it more efficient somehow? not sure tho 59 | fg_featurizer = FunctionalGroupFeaturizer(args) 60 | fg_features = fg_featurizer.featurize(mol) 61 | features += np.array(fg_features).sum(axis=0).tolist() 62 | features = np.clip(np.nan_to_num(np.array(features)), -1e2, 1e2) 63 | 64 | return features 65 | -------------------------------------------------------------------------------- /chemprop/features/rdkit_normalized_features.py: -------------------------------------------------------------------------------- 1 | import logging 2 | from rdkit import Chem 3 | 4 | try: 5 | from descriptastorus.descriptors import rdNormalizedDescriptors 6 | generator = rdNormalizedDescriptors.RDKit2DNormalized() 7 | 8 | def rdkit_2d_normalized_features(smiles: str): 9 | # the first element is true/false if the mol was properly computed 10 | if type(smiles) == str: 11 | return generator.process(smiles)[1:] 12 | 13 | else: 14 | # this is a bit of a waste, but the desciptastorus API is smiles 15 | # based for normalization purposes 16 | return generator.process(Chem.MolToSmiles(smiles, isomericSmiles=True))[1:] 17 | 18 | 19 | except ImportError: 20 | logging.getLogger(__name__).warning("descriptastorus is not available, normalized descriptors are not available") 21 | rdkit_2d_normalized_features = None 22 | 23 | 24 | 25 | 26 | 27 | -------------------------------------------------------------------------------- /chemprop/features/utils.py: -------------------------------------------------------------------------------- 1 | from argparse import Namespace 2 | from functools import partial 3 | import os 4 | import pickle 5 | from typing import Callable, List, Union 6 | import logging 7 | 8 | import numpy as np 9 | from rdkit import Chem 10 | 11 | from .descriptors import mordred_features 12 | from .morgan_fingerprint import morgan_fingerprint 13 | from .rdkit_features import rdkit_2d_features 14 | 15 | 16 | def load_features(path: str) -> List[np.ndarray]: 17 | """ 18 | Loads features saved as a .pckl file or as a directory of .pckl files. 19 | 20 | If path is a directory, assumes features are saved in files named 0.pckl, 1.pckl, ... 21 | 22 | :param path: Path to a .pckl file or a directory of .pckl files named as above. 23 | :return: A list of numpy arrays containing the features. 24 | """ 25 | if os.path.isfile(path): 26 | with open(path, 'rb') as f: 27 | features = pickle.load(f) 28 | features = [np.squeeze(np.array(feat.todense())) for feat in features] 29 | else: 30 | features = [] 31 | features_num = 0 32 | features_path = os.path.join(path, f'{features_num}.pckl') 33 | 34 | while os.path.exists(features_path): 35 | with open(features_path, 'rb') as f: 36 | feats = pickle.load(f) 37 | features.extend([np.squeeze(np.array(feat.todense())) for feat in feats]) 38 | 39 | features_num += 1 40 | features_path = os.path.join(path, f'{features_num}.pckl') 41 | 42 | return features 43 | 44 | 45 | def get_features_func(features_generator: str, 46 | args: Namespace = None) -> Union[Callable[[Chem.Mol], np.ndarray], 47 | partial]: 48 | if features_generator == 'morgan': 49 | return partial(morgan_fingerprint, use_counts=False) 50 | 51 | if features_generator == 'morgan_count': 52 | return partial(morgan_fingerprint, use_counts=True) 53 | 54 | if features_generator == 'rdkit_2d': 55 | assert args is not None 56 | assert hasattr(args, 'functional_group_smarts') # TODO: handle this in a better way 57 | return partial(rdkit_2d_features, args=args) 58 | 59 | if features_generator == "rdkit_2d_normalized": 60 | from .rdkit_normalized_features import rdkit_2d_normalized_features 61 | if rdkit_2d_normalized_features is None: 62 | logging.getLogger(__name__).warning("Please install descriptastorus for normalized descriptors") 63 | raise ValueError(f'feature_generator type "{features_generator}" not installed') 64 | 65 | return rdkit_2d_normalized_features 66 | 67 | if features_generator == 'mordred': 68 | return mordred_features 69 | 70 | raise ValueError(f'features_generator type "{features_generator}" not supported.') 71 | -------------------------------------------------------------------------------- /chemprop/models/__init__.py: -------------------------------------------------------------------------------- 1 | from .model import build_model 2 | -------------------------------------------------------------------------------- /chemprop/models/learned_kernel.py: -------------------------------------------------------------------------------- 1 | from argparse import Namespace 2 | 3 | import torch 4 | import torch.nn as nn 5 | 6 | 7 | class LearnedKernel(nn.Module): 8 | def __init__(self, args: Namespace): 9 | super(LearnedKernel, self).__init__() 10 | self.A = nn.Linear(args.ffn_hidden_size, args.ffn_hidden_size) 11 | 12 | def forward(self, encodings: torch.Tensor): 13 | # encodings is num_pairs x 2 x ffn hidden size 14 | return (self.A(encodings[:, 1, :].squeeze(1)) * encodings[:, 0, :].squeeze(1)).sum(dim=1, keepdim=True) 15 | -------------------------------------------------------------------------------- /chemprop/train/__init__.py: -------------------------------------------------------------------------------- 1 | from .cross_validate import cross_validate 2 | from .evaluate import evaluate, evaluate_predictions 3 | from .make_predictions import make_predictions 4 | from .predict import predict 5 | from .run_training import run_training 6 | from .train import train 7 | -------------------------------------------------------------------------------- /chemprop/train/cross_validate.py: -------------------------------------------------------------------------------- 1 | from argparse import Namespace 2 | from logging import Logger 3 | import os 4 | from typing import Tuple 5 | 6 | import numpy as np 7 | import shutil 8 | 9 | from chemprop.data.utils import get_task_names, get_desired_labels 10 | from .run_training import run_training 11 | 12 | 13 | def cross_validate(args: Namespace, logger: Logger = None) -> Tuple[float, float]: 14 | """k-fold cross validation""" 15 | info = logger.info if logger is not None else print 16 | 17 | # Initialize relevant variables 18 | init_seed = args.seed 19 | save_dir = args.save_dir 20 | task_names = get_task_names(args.data_path) 21 | desired_labels = get_desired_labels(args, task_names) 22 | 23 | # Run training on different random seeds for each fold 24 | all_scores = [] 25 | for fold_num in range(args.num_folds): 26 | info(f'Fold {fold_num}') 27 | args.seed = init_seed + fold_num 28 | args.save_dir = os.path.join(save_dir, f'fold_{fold_num}') 29 | os.makedirs(args.save_dir, exist_ok=True) 30 | model_scores = run_training(args, logger) 31 | all_scores.append(model_scores) 32 | all_scores = np.array(all_scores) 33 | 34 | # Report results 35 | info(f'{args.num_folds}-fold cross validation') 36 | 37 | # Report scores for each fold 38 | for fold_num, scores in enumerate(all_scores): 39 | info(f'Seed {init_seed + fold_num} ==> test {args.metric} = {np.nanmean(scores):.6f}') 40 | 41 | if args.show_individual_scores: 42 | for task_name, score in zip(task_names, scores): 43 | if task_name in desired_labels: 44 | info(f'Seed {init_seed + fold_num} ==> test {task_name} {args.metric} = {score:.6f}') 45 | 46 | # Report scores across models 47 | avg_scores = np.nanmean(all_scores, axis=1) # average score for each model across tasks 48 | mean_score, std_score = np.nanmean(avg_scores), np.nanstd(avg_scores) 49 | info(f'Overall test {args.metric} = {mean_score:.6f} +/- {std_score:.6f}') 50 | 51 | if args.show_individual_scores: 52 | for task_num, task_name in enumerate(task_names): 53 | if task_name in desired_labels: 54 | info(f'Overall test {task_name} {args.metric} = ' 55 | f'{np.nanmean(all_scores[:, task_num]):.6f} +/- {np.nanstd(all_scores[:, task_num]):.6f}') 56 | 57 | if args.num_chunks > 1: 58 | shutil.rmtree(args.chunk_temp_dir) 59 | 60 | return mean_score, std_score 61 | -------------------------------------------------------------------------------- /chemprop/train/make_predictions.py: -------------------------------------------------------------------------------- 1 | from argparse import Namespace 2 | import csv 3 | from typing import List, Optional 4 | 5 | import numpy as np 6 | import torch 7 | from tqdm import tqdm 8 | from rdkit import Chem 9 | 10 | from .predict import predict 11 | from chemprop.data import MoleculeDataset 12 | from chemprop.data.utils import get_data, get_data_from_smiles 13 | from chemprop.utils import load_args, load_checkpoint, load_scalers 14 | 15 | 16 | def make_predictions(args: Namespace, smiles: List[str] = None) -> List[Optional[List[float]]]: 17 | """ 18 | Makes predictions. If smiles is provided, makes predictions on smiles. Otherwise makes predictions on args.test_data. 19 | 20 | :param args: Arguments. 21 | :param smiles: Smiles to make predictions on. 22 | :return: A list of lists of target predictions. 23 | """ 24 | if args.gpu is not None: 25 | torch.cuda.set_device(args.gpu) 26 | 27 | print('Loading training args') 28 | scaler, features_scaler = load_scalers(args.checkpoint_paths[0]) 29 | train_args = load_args(args.checkpoint_paths[0]) 30 | 31 | # Update args with training arguments 32 | for key, value in vars(train_args).items(): 33 | if not hasattr(args, key): 34 | setattr(args, key, value) 35 | 36 | print('Loading data') 37 | if smiles is not None: 38 | test_data = get_data_from_smiles(smiles=smiles, skip_invalid_smiles=False) 39 | else: 40 | test_data = get_data(path=args.test_path, args=args, use_compound_names=args.compound_names, skip_invalid_smiles=False) 41 | 42 | print('Validating SMILES') 43 | valid_indices = [i for i in range(len(test_data)) if test_data[i].mol is not None] 44 | full_data = test_data 45 | test_data = MoleculeDataset([test_data[i] for i in valid_indices]) 46 | 47 | # Edge case if empty list of smiles is provided 48 | if len(test_data) == 0: 49 | return [None] * len(full_data) 50 | 51 | test_smiles = test_data.smiles() 52 | 53 | if args.compound_names: 54 | compound_names = test_data.compound_names() 55 | print(f'Test size = {len(test_data):,}') 56 | 57 | # Normalize features 58 | if train_args.features_scaling: 59 | test_data.normalize_features(features_scaler) 60 | 61 | # Predict with each model individually and sum predictions 62 | sum_preds = np.zeros((len(test_data), args.num_tasks)) 63 | print(f'Predicting with an ensemble of {len(args.checkpoint_paths)} models') 64 | for checkpoint_path in tqdm(args.checkpoint_paths, total=len(args.checkpoint_paths)): 65 | # Load model 66 | model = load_checkpoint(checkpoint_path, cuda=args.cuda) 67 | model_preds = predict( 68 | model=model, 69 | data=test_data, 70 | args=args, 71 | scaler=scaler 72 | ) 73 | sum_preds += np.array(model_preds) 74 | 75 | # Ensemble predictions 76 | avg_preds = sum_preds / args.ensemble_size 77 | avg_preds = avg_preds.tolist() 78 | 79 | # Save predictions 80 | assert len(test_data) == len(avg_preds) 81 | print(f'Saving predictions to {args.preds_path}') 82 | 83 | # Put Nones for invalid smiles 84 | full_preds = [None] * len(full_data) 85 | for i, si in enumerate(valid_indices): 86 | full_preds[si] = avg_preds[i] 87 | avg_preds = full_preds 88 | test_smiles = full_data.smiles() 89 | 90 | # Write predictions 91 | with open(args.preds_path, 'w') as f: 92 | writer = csv.writer(f) 93 | 94 | header = [] 95 | if args.write_smiles: 96 | header.append('smiles') 97 | if args.compound_names: 98 | header.append('compound_names') 99 | 100 | header.extend(args.task_names) 101 | writer.writerow(header) 102 | 103 | for i in range(len(avg_preds)): 104 | row = [] 105 | 106 | if args.write_smiles: 107 | row.append(test_smiles[i]) 108 | if args.compound_names: 109 | row.append(compound_names[i]) 110 | 111 | if avg_preds[i] is not None: 112 | row.extend(avg_preds[i]) 113 | else: 114 | row.extend([''] * args.num_tasks) 115 | 116 | writer.writerow(row) 117 | 118 | return avg_preds 119 | -------------------------------------------------------------------------------- /data/all_smiles.csv.gz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/data/all_smiles.csv.gz -------------------------------------------------------------------------------- /data/hiv.csv.gz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/data/hiv.csv.gz -------------------------------------------------------------------------------- /data/muv.csv.gz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/data/muv.csv.gz -------------------------------------------------------------------------------- /data/pcba.csv.gz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/data/pcba.csv.gz -------------------------------------------------------------------------------- /data/qm7.csv.gz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/data/qm7.csv.gz -------------------------------------------------------------------------------- /data/qm8.csv.gz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/data/qm8.csv.gz -------------------------------------------------------------------------------- /data/qm9.csv.gz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/data/qm9.csv.gz -------------------------------------------------------------------------------- /data/sanitize.py: -------------------------------------------------------------------------------- 1 | from argparse import ArgumentParser 2 | import csv 3 | from rdkit import Chem 4 | 5 | 6 | def sanitize(data_path: str, save_path): 7 | with open(data_path) as f: 8 | reader = csv.reader(f) 9 | header = next(reader) 10 | lines = [line for line in reader if line[0] != '' and Chem.MolFromSmiles(line[0]) is not None] 11 | 12 | with open(save_path) as f: 13 | writer = csv.writer(f) 14 | writer.writerow(header) 15 | for line in lines: 16 | writer.writerow(line) 17 | 18 | 19 | if __name__ == '__main__': 20 | parser = ArgumentParser() 21 | parser.add_argument('--data_path', type=str, required=True, help='Data CSV to sanitize') 22 | parser.add_argument('--save_path', type=str, required=True, help='Path to CSV where sanitized data will be saved') 23 | args = parser.parse_args() 24 | 25 | sanitize(args.data_path, args.save_path) 26 | -------------------------------------------------------------------------------- /data/toxcast.csv.gz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/data/toxcast.csv.gz -------------------------------------------------------------------------------- /distributions/chembl/chembl_labels_log.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/chembl/chembl_labels_log.png -------------------------------------------------------------------------------- /distributions/delaney/delaney_logp.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/delaney/delaney_logp.png -------------------------------------------------------------------------------- /distributions/freesolv/freesolv_freesolv.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/freesolv/freesolv_freesolv.png -------------------------------------------------------------------------------- /distributions/lipo/lipo_lipo.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/lipo/lipo_lipo.png -------------------------------------------------------------------------------- /distributions/qm8/qm8_E1-CAM.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm8/qm8_E1-CAM.png -------------------------------------------------------------------------------- /distributions/qm8/qm8_E1-CC2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm8/qm8_E1-CC2.png -------------------------------------------------------------------------------- /distributions/qm8/qm8_E1-PBE0.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm8/qm8_E1-PBE0.png -------------------------------------------------------------------------------- /distributions/qm8/qm8_E2-CAM.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm8/qm8_E2-CAM.png -------------------------------------------------------------------------------- /distributions/qm8/qm8_E2-CC2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm8/qm8_E2-CC2.png -------------------------------------------------------------------------------- /distributions/qm8/qm8_E2-PBE0.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm8/qm8_E2-PBE0.png -------------------------------------------------------------------------------- /distributions/qm8/qm8_f1-CAM.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm8/qm8_f1-CAM.png -------------------------------------------------------------------------------- /distributions/qm8/qm8_f1-CC2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm8/qm8_f1-CC2.png -------------------------------------------------------------------------------- /distributions/qm8/qm8_f1-PBE0.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm8/qm8_f1-PBE0.png -------------------------------------------------------------------------------- /distributions/qm8/qm8_f2-CAM.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm8/qm8_f2-CAM.png -------------------------------------------------------------------------------- /distributions/qm8/qm8_f2-CC2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm8/qm8_f2-CC2.png -------------------------------------------------------------------------------- /distributions/qm8/qm8_f2-PBE0.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm8/qm8_f2-PBE0.png -------------------------------------------------------------------------------- /distributions/qm9/qm9_alpha.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm9/qm9_alpha.png -------------------------------------------------------------------------------- /distributions/qm9/qm9_cv.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm9/qm9_cv.png -------------------------------------------------------------------------------- /distributions/qm9/qm9_g298.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm9/qm9_g298.png -------------------------------------------------------------------------------- /distributions/qm9/qm9_g298_atom.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm9/qm9_g298_atom.png -------------------------------------------------------------------------------- /distributions/qm9/qm9_gap.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm9/qm9_gap.png -------------------------------------------------------------------------------- /distributions/qm9/qm9_h298.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm9/qm9_h298.png -------------------------------------------------------------------------------- /distributions/qm9/qm9_h298_atom.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm9/qm9_h298_atom.png -------------------------------------------------------------------------------- /distributions/qm9/qm9_homo.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm9/qm9_homo.png -------------------------------------------------------------------------------- /distributions/qm9/qm9_lumo.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm9/qm9_lumo.png -------------------------------------------------------------------------------- /distributions/qm9/qm9_mu.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm9/qm9_mu.png -------------------------------------------------------------------------------- /distributions/qm9/qm9_r2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm9/qm9_r2.png -------------------------------------------------------------------------------- /distributions/qm9/qm9_u0.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm9/qm9_u0.png -------------------------------------------------------------------------------- /distributions/qm9/qm9_u0_atom.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm9/qm9_u0_atom.png -------------------------------------------------------------------------------- /distributions/qm9/qm9_u298.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm9/qm9_u298.png -------------------------------------------------------------------------------- /distributions/qm9/qm9_u298_atom.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm9/qm9_u298_atom.png -------------------------------------------------------------------------------- /distributions/qm9/qm9_zpve.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/distributions/qm9/qm9_zpve.png -------------------------------------------------------------------------------- /hyperparameter_optimization.py: -------------------------------------------------------------------------------- 1 | from argparse import ArgumentParser, Namespace 2 | from copy import deepcopy 3 | import json 4 | import os 5 | from typing import Dict, Union 6 | 7 | from hyperopt import fmin, hp, tpe 8 | import numpy as np 9 | 10 | from chemprop.models import build_model 11 | from chemprop.nn_utils import param_count 12 | from chemprop.parsing import add_train_args, modify_train_args 13 | from chemprop.train import cross_validate 14 | from model_comparison import create_logger, create_train_logger 15 | 16 | 17 | SPACE = { 18 | 'hidden_size': hp.quniform('hidden_size', low=300, high=2400, q=100), 19 | 'depth': hp.quniform('depth', low=2, high=6, q=1), 20 | 'dropout': hp.quniform('dropout', low=0.0, high=0.4, q=0.05), 21 | 'ffn_num_layers': hp.quniform('ffn_num_layers', low=1, high=3, q=1) 22 | } 23 | INT_KEYS = ['hidden_size', 'depth', 'ffn_num_layers'] 24 | 25 | TRAIN_LOGGER = create_train_logger() 26 | 27 | 28 | def grid_search(args: Namespace): 29 | # Create logger for dataset 30 | logger = create_logger(name='hyperparameter_optimization', save_path=args.log_path) 31 | 32 | # Run grid search 33 | results = [] 34 | 35 | # Define hyperparameter optimization 36 | def objective(hyperparams: Dict[str, Union[int, float]]) -> float: 37 | # Convert hyperparams from float to int when necessary 38 | for key in INT_KEYS: 39 | hyperparams[key] = int(hyperparams[key]) 40 | 41 | # Update args with hyperparams 42 | hyper_args = deepcopy(args) 43 | if args.save_dir is not None: 44 | folder_name = '_'.join([f'{key}_{value}' if key in INT_KEYS else f'{key}_{value}' for key, value in hyperparams.items()]) 45 | hyper_args.save_dir = os.path.join(hyper_args.save_dir, folder_name) 46 | for key, value in hyperparams.items(): 47 | setattr(hyper_args, key, value) 48 | 49 | # Record hyperparameters 50 | logger.info(hyperparams) 51 | 52 | # Cross validate 53 | mean_score, std_score = cross_validate(hyper_args, TRAIN_LOGGER) 54 | 55 | # Record results 56 | temp_model = build_model(hyper_args) 57 | num_params = param_count(temp_model) 58 | logger.info(f'num params: {num_params:,}') 59 | logger.info(f'{mean_score} +/- {std_score} {hyper_args.metric}') 60 | 61 | results.append({ 62 | 'mean_score': mean_score, 63 | 'std_score': std_score, 64 | 'hyperparams': hyperparams, 65 | 'num_params': num_params 66 | }) 67 | 68 | # Deal with nan 69 | if np.isnan(mean_score): 70 | if hyper_args.dataset_type == 'classification': 71 | mean_score = 0 72 | else: 73 | raise ValueError('Can\'t handle nan score for non-classification dataset.') 74 | 75 | return (1 if hyper_args.minimize_score else -1) * mean_score 76 | 77 | fmin(objective, SPACE, algo=tpe.suggest, max_evals=args.num_iters) 78 | 79 | # Report best result 80 | results = [result for result in results if not np.isnan(result['mean_score'])] 81 | best_result = min(results, key=lambda result: (1 if args.minimize_score else -1) * result['mean_score']) 82 | logger.info('best') 83 | logger.info(best_result['hyperparams']) 84 | logger.info(f'num params: {best_result["num_params"]:,}') 85 | logger.info(f'{best_result["mean_score"]} +/- {best_result["std_score"]} {args.metric}') 86 | 87 | # Save best hyperparameter settings as JSON config file 88 | with open(args.config_save_path, 'w') as f: 89 | json.dump(best_result['hyperparams'], f, indent=4, sort_keys=True) 90 | 91 | 92 | if __name__ == '__main__': 93 | parser = ArgumentParser() 94 | add_train_args(parser) 95 | parser.add_argument('--num_iters', type=int, default=20, 96 | help='Number of hyperparameter choices to try') 97 | parser.add_argument('--config_save_path', type=str, required=True, 98 | help='Path to .json file where best hyperparameter settings will be written') 99 | parser.add_argument('--log_path', type=str, 100 | help='(Optional) Path to .log file where all results of the hyperparameter optimization will be written') 101 | args = parser.parse_args() 102 | modify_train_args(args) 103 | 104 | grid_search(args) 105 | -------------------------------------------------------------------------------- /predict.py: -------------------------------------------------------------------------------- 1 | from chemprop.parsing import parse_predict_args 2 | from chemprop.train import make_predictions 3 | 4 | if __name__ == '__main__': 5 | args = parse_predict_args() 6 | make_predictions(args) 7 | -------------------------------------------------------------------------------- /random_forest.py: -------------------------------------------------------------------------------- 1 | from argparse import ArgumentParser 2 | import logging 3 | 4 | from chemprop.random_forest import cross_validate_random_forest 5 | from chemprop.utils import set_logger 6 | 7 | logger = logging.getLogger('random_forest') 8 | logger.setLevel(logging.DEBUG) 9 | logger.propagate = False 10 | 11 | 12 | if __name__ == '__main__': 13 | parser = ArgumentParser() 14 | parser.add_argument('--data_path', type=str, required=True, 15 | help='Path to data CSV') 16 | parser.add_argument('--dataset_type', type=str, required=True, 17 | choices=['regression', 'classification'], 18 | help='Dataset type') 19 | parser.add_argument('--metric', type=str, 20 | choices=['auc', 'prc-auc', 'rmse', 'mae'], 21 | help='Metric to use during evaluation.') 22 | parser.add_argument('--split_type', type=str, default='random', 23 | choices=['random', 'scaffold_balanced'], 24 | help='Split type') 25 | parser.add_argument('--class_weight', type=str, 26 | choices=['balanced'], 27 | help='How to weight classes (None means no class balance)') 28 | parser.add_argument('--single_task', action='store_true', default=False, 29 | help='Whether to run each task separately (needed when dataset has null entries)') 30 | parser.add_argument('--num_folds', type=int, default=1, 31 | help='Number of folds of cross validation') 32 | parser.add_argument('--seed', type=int, default=0, 33 | help='Random seed') 34 | parser.add_argument('--radius', type=int, default=2, 35 | help='Morgan fingerprint radius') 36 | parser.add_argument('--num_bits', type=int, default=2048, 37 | help='Number of bits in morgan fingerprint') 38 | parser.add_argument('--num_trees', type=int, default=500, 39 | help='Number of random forest trees') 40 | parser.add_argument('--quiet', action='store_true', default=False, 41 | help='Control verbosity level') 42 | args = parser.parse_args() 43 | 44 | set_logger(logger, quiet=args.quiet) 45 | 46 | if args.metric is None: 47 | if args.dataset_type == 'regression': 48 | args.metric = 'rmse' 49 | elif args.dataset_type == 'classification': 50 | args.metric = 'auc' 51 | else: 52 | raise ValueError(f'Default metric not supported for dataset_type "{args.dataset_type}"') 53 | 54 | cross_validate_random_forest(args, logger) 55 | -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | flask 2 | hyperopt 3 | matplotlib 4 | mordred 5 | networkx 6 | numpy 7 | python-ternary 8 | scikit-learn 9 | scipy 10 | tensorboardX 11 | tqdm 12 | git+https://github.com/bp-kelley/descriptastorus 13 | -------------------------------------------------------------------------------- /scripts/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/scripts/__init__.py -------------------------------------------------------------------------------- /scripts/avg_dups.py: -------------------------------------------------------------------------------- 1 | from argparse import ArgumentParser 2 | import sys 3 | sys.path.append('../') 4 | 5 | from chemprop.data_processing import average_duplicates 6 | 7 | if __name__ == '__main__': 8 | parser = ArgumentParser() 9 | parser.add_argument('--data_path', type=str, 10 | help='Path to data CSV file') 11 | parser.add_argument('--save_path', type=str, 12 | help='Path where average data CSV file will be saved') 13 | args = parser.parse_args() 14 | 15 | average_duplicates(args) 16 | -------------------------------------------------------------------------------- /scripts/filter_by_scaffold.py: -------------------------------------------------------------------------------- 1 | from argparse import ArgumentParser, Namespace 2 | from collections import namedtuple 3 | import sys 4 | sys.path.append('../') 5 | from typing import List 6 | 7 | from tqdm import tqdm 8 | 9 | from chemprop.data.scaffold import generate_scaffold 10 | 11 | 12 | Datapoint = namedtuple('Datapoint', ['smiles', 'line']) 13 | 14 | 15 | def get_header(path: str) -> str: 16 | with open(path) as f: 17 | header = f.readline() 18 | 19 | return header 20 | 21 | 22 | def get_data(path: str) -> List[Datapoint]: 23 | with open(path) as f: 24 | f.readline() # skip header 25 | data = [] 26 | for line in f.readlines(): 27 | smiles = line[:line.index(',')] 28 | data.append(Datapoint(smiles=smiles, line=line)) 29 | 30 | return data 31 | 32 | 33 | def filter_by_scaffold(args: Namespace): 34 | print('Loading data') 35 | header = get_header(args.data_path) 36 | data = get_data(path=args.data_path) 37 | scaffold_data = get_data(path=args.scaffold_data_path) 38 | 39 | print('Generating scaffolds') 40 | smiles_to_scaffold = {d.smiles: generate_scaffold(d.smiles) for d in tqdm(data, total=len(data))} 41 | scaffolds_to_keep = {generate_scaffold(d.smiles) for d in tqdm(scaffold_data, total=len(scaffold_data))} 42 | 43 | print('Filtering data') 44 | filtered_data = [d for d in data if smiles_to_scaffold[d.smiles] in scaffolds_to_keep] 45 | 46 | print(f'Filtered data from {len(data):,} to {len(filtered_data):,} molecules') 47 | 48 | print('Saving data') 49 | with open(args.save_path, 'w') as f: 50 | f.write(header) 51 | for d in filtered_data: 52 | f.write(d.line) 53 | 54 | 55 | if __name__ == '__main__': 56 | parser = ArgumentParser() 57 | parser.add_argument('--data_path', type=str, required=True, 58 | help='Path to dataset which will be limited to those molecules sharing' 59 | 'a scaffold with a molecule in the scaffold_data_path dataset') 60 | parser.add_argument('--scaffold_data_path', type=str, required=True, 61 | help='Path to the dataset whose scaffolds will be used to limit the' 62 | 'molecules in data_path') 63 | parser.add_argument('--save_path', type=str, required=True, 64 | help='Path where the filtered version of data_path will be saved') 65 | args = parser.parse_args() 66 | 67 | filter_by_scaffold(args) 68 | -------------------------------------------------------------------------------- /scripts/overlap.py: -------------------------------------------------------------------------------- 1 | from argparse import ArgumentParser 2 | import csv 3 | import sys 4 | sys.path.append('../') 5 | 6 | from chemprop.data.utils import get_data 7 | 8 | if __name__ == '__main__': 9 | parser = ArgumentParser() 10 | parser.add_argument('--data_path_1', type=str, required=True, 11 | help='Path to first data CSV file') 12 | parser.add_argument('--data_path_2', type=str, required=True, 13 | help='Path to second data CSV file') 14 | parser.add_argument('--compound_names_1', action='store_true', default=False, 15 | help='Whether data_path_1 has compound names in addition to smiles') 16 | parser.add_argument('--compound_names_2', action='store_true', default=False, 17 | help='Whether data_path_2 has compound names in addition to smiles') 18 | parser.add_argument('--save_intersection_path', type=str, default=None, 19 | help='Path to save intersection at; labeled with data_path 1 header') 20 | parser.add_argument('--save_difference_path', type=str, default=None, 21 | help='Path to save molecules in dataset 1 that are not in dataset 2; labeled with data_path 1 header') 22 | args = parser.parse_args() 23 | 24 | data_1 = get_data(path=args.data_path_1, use_compound_names=args.compound_names_1) 25 | data_2 = get_data(path=args.data_path_2, use_compound_names=args.compound_names_2) 26 | 27 | smiles1 = set(data_1.smiles()) 28 | smiles2 = set(data_2.smiles()) 29 | size_1, size_2 = len(smiles1), len(smiles2) 30 | intersection = smiles1.intersection(smiles2) 31 | size_intersect = len(intersection) 32 | print(f'Size of dataset 1: {size_1}') 33 | print(f'Size of dataset 2: {size_2}') 34 | print(f'Size of intersection: {size_intersect}') 35 | print(f'Size of intersection as frac of dataset 1: {size_intersect / size_1}') 36 | print(f'Size of intersection as frac of dataset 2: {size_intersect / size_2}') 37 | 38 | if args.save_intersection_path is not None: 39 | with open(args.data_path_1, 'r') as rf, open(args.save_intersection_path, 'w') as wf: 40 | reader, writer = csv.reader(rf), csv.writer(wf) 41 | header = next(reader) 42 | writer.writerow(header) 43 | for line in reader: 44 | if line[0] in intersection: 45 | writer.writerow(line) 46 | 47 | if args.save_difference_path is not None: 48 | with open(args.data_path_1, 'r') as rf, open(args.save_difference_path, 'w') as wf: 49 | reader, writer = csv.reader(rf), csv.writer(wf) 50 | header = next(reader) 51 | writer.writerow(header) 52 | for line in reader(): 53 | if line[0] not in intersection: 54 | writer.writerow(line) 55 | -------------------------------------------------------------------------------- /scripts/plot_distribution.py: -------------------------------------------------------------------------------- 1 | from argparse import ArgumentParser 2 | import os 3 | import sys 4 | sys.path.append('../') 5 | 6 | from chemprop.data_processing import plot_distribution 7 | 8 | 9 | if __name__ == '__main__': 10 | parser = ArgumentParser() 11 | parser.add_argument('--data_path', type=str, 12 | help='Path to data CSV file') 13 | parser.add_argument('--save_dir', type=str, 14 | help='Directory where plot PNGs will be saved') 15 | parser.add_argument('--bins', type=int, default=50, 16 | help='Number of bins in histogram.') 17 | args = parser.parse_args() 18 | 19 | os.makedirs(args.save_dir, exist_ok=True) 20 | 21 | plot_distribution(args.data_path, args.save_dir, args.bins) 22 | -------------------------------------------------------------------------------- /scripts/resplit_data.py: -------------------------------------------------------------------------------- 1 | from argparse import ArgumentParser 2 | import os 3 | import sys 4 | sys.path.append('../') 5 | 6 | from chemprop.data_processing import resplit 7 | 8 | 9 | if __name__ == '__main__': 10 | parser = ArgumentParser() 11 | parser.add_argument('--train_path', type=str, required=True, 12 | help='Path to CSV file containing training data') 13 | parser.add_argument('--val_path', type=str, required=True, 14 | help='Path to CSV file containing val data') 15 | parser.add_argument('--train_save', type=str, required=True, 16 | help='Path to CSV file for new train data') 17 | parser.add_argument('--val_save', type=str, required=True, 18 | help='Path to CSV file for new val data') 19 | parser.add_argument('--val_frac', type=float, default=0.2, 20 | help='frac of data to use for validation') 21 | args = parser.parse_args() 22 | 23 | # Create directory for save_path 24 | for path in [args.train_save, args.val_save]: 25 | save_dir = os.path.dirname(path) 26 | if save_dir != '': 27 | os.makedirs(save_dir, exist_ok=True) 28 | 29 | resplit(args) 30 | -------------------------------------------------------------------------------- /scripts/similarity.py: -------------------------------------------------------------------------------- 1 | from argparse import ArgumentParser 2 | import sys 3 | sys.path.append('../') 4 | 5 | from chemprop.data.utils import get_data 6 | from chemprop.data import morgan_similarity, scaffold_similarity 7 | 8 | 9 | if __name__ == '__main__': 10 | parser = ArgumentParser() 11 | parser.add_argument('--data_path_1', type=str, required=True, 12 | help='Path to first data CSV file') 13 | parser.add_argument('--data_path_2', type=str, required=True, 14 | help='Path to second data CSV file') 15 | parser.add_argument('--compound_names_1', action='store_true', default=False, 16 | help='Whether data_path_1 has compound names in addition to smiles') 17 | parser.add_argument('--compound_names_2', action='store_true', default=False, 18 | help='Whether data_path_2 has compound names in addition to smiles') 19 | parser.add_argument('--similarity_measure', type=str, required=True, choices=['scaffold', 'morgan'], 20 | help='Similarity measure to use to compare the two datasets') 21 | parser.add_argument('--radius', type=int, default=3, 22 | help='Radius of Morgan fingerprint') 23 | parser.add_argument('--sample_rate', type=float, default=1.0, 24 | help='Rate at which to sample pairs of molecules for Morgan similarity (to reduce time)') 25 | args = parser.parse_args() 26 | 27 | data_1 = get_data(path=args.data_path_1, use_compound_names=args.compound_names_1) 28 | data_2 = get_data(path=args.data_path_2, use_compound_names=args.compound_names_2) 29 | 30 | if args.similarity_measure == 'scaffold': 31 | scaffold_similarity(data_1.smiles(), data_2.smiles()) 32 | elif args.similarity_measure == 'morgan': 33 | morgan_similarity(data_1.smiles(), data_2.smiles(), args.radius, args.sample_rate) 34 | else: 35 | raise ValueError(f'Similarity measure "{args.similarity_measure}" not supported.') 36 | -------------------------------------------------------------------------------- /scripts/visualize_encoding_property_space.py: -------------------------------------------------------------------------------- 1 | from argparse import ArgumentParser, Namespace 2 | import os 3 | import random 4 | import sys 5 | sys.path.append('../') 6 | from typing import List 7 | 8 | import numpy as np 9 | import ternary 10 | from tqdm import trange 11 | 12 | from chemprop.data import MoleculeDataset 13 | from chemprop.data.utils import get_data 14 | from chemprop.utils import load_checkpoint, load_scalers 15 | 16 | 17 | def visualize_encoding_property_space(args: Namespace): 18 | # Load data 19 | data = get_data(path=args.data_path) 20 | 21 | # Sort according to similarity measure 22 | if args.similarity_measure == 'property': 23 | data.sort(key=lambda d: d.targets[args.task_index]) 24 | elif args.similarity_measure == 'random': 25 | data.shuffle(args.seed) 26 | else: 27 | raise ValueError(f'similarity_measure "{args.similarity_measure}" not supported or not implemented yet.') 28 | 29 | # Load model and scalers 30 | model = load_checkpoint(args.checkpoint_path) 31 | scaler, features_scaler = load_scalers(args.checkpoint_path) 32 | data.normalize_features(features_scaler) 33 | 34 | # Random seed 35 | if args.seed is not None: 36 | random.seed(args.seed) 37 | 38 | # Generate visualizations 39 | for i in trange(args.num_examples): 40 | # Get random three molecules with similar properties 41 | index = random.randint(1, len(data) - 2) 42 | molecules = MoleculeDataset(data[index - 1:index + 2]) 43 | molecule_targets = [t[args.task_index] for t in molecules.targets()] 44 | 45 | # Encode three molecules 46 | molecule_encodings = model.encoder(molecules.smiles()) 47 | 48 | # Define interpolation 49 | def predict_property(point: List[int]) -> float: 50 | # Return true value on endpoints of triangle 51 | argmax = np.argmax(point) 52 | if point[argmax] == 1: 53 | return molecule_targets[argmax] 54 | 55 | # Interpolate and predict task value 56 | encoding = sum(point[j] * molecule_encodings[j] for j in range(len(molecule_encodings))) 57 | pred = model.ffn(encoding).data.cpu().numpy() 58 | pred = scaler.inverse_transform(pred) 59 | pred = pred.item() 60 | 61 | return pred 62 | 63 | # Create visualization 64 | scale = 20 65 | fontsize = 6 66 | 67 | figure, tax = ternary.figure(scale=scale) 68 | tax.heatmapf(predict_property, boundary=True, style="hexagonal") 69 | tax.set_title("Property Prediction") 70 | tax.right_axis_label(f'{molecules[0].smiles} ({molecules[0].targets[args.task_index]:.6f}) -->', 71 | fontsize=fontsize) 72 | tax.left_axis_label(f'{molecules[1].smiles} ({molecules[1].targets[args.task_index]:.6f}) -->', 73 | fontsize=fontsize) 74 | tax.bottom_axis_label(f'<-- {molecules[2].smiles} ({molecules[2].targets[args.task_index]:.6f})', 75 | fontsize=fontsize) 76 | 77 | tax.savefig(os.path.join(args.save_dir, f'{i}.png')) 78 | 79 | 80 | if __name__ == '__main__': 81 | parser = ArgumentParser() 82 | parser.add_argument('--data_path', type=str, required=True, 83 | help='Path to regression dataset .csv') 84 | parser.add_argument('--checkpoint_path', type=str, required=True, 85 | help='Path to a model checkpoint .pt file') 86 | parser.add_argument('--similarity_measure', type=str, default='random', 87 | choices=['random', 'random'], 88 | help='Similarity measure to use when choosing the three molecules for each visualization') 89 | parser.add_argument('--task_index', type=int, default=0, 90 | help='Index of the task (property) in the dataset to use') 91 | parser.add_argument('--num_examples', type=int, default=10, 92 | help='Number of visualizations to generate') 93 | parser.add_argument('--save_dir', type=str, required=True, 94 | help='Directory where visualizations will be saved') 95 | parser.add_argument('--seed', type=int, default=None, 96 | help='Random seed for choosing three similar molecules') 97 | args = parser.parse_args() 98 | 99 | os.makedirs(args.save_dir, exist_ok=True) 100 | 101 | visualize_encoding_property_space(args) 102 | -------------------------------------------------------------------------------- /scripts/viz_attention.py: -------------------------------------------------------------------------------- 1 | from argparse import ArgumentParser, Namespace 2 | import os 3 | import sys 4 | sys.path.append('../') 5 | 6 | import torch 7 | from tqdm import trange 8 | 9 | from chemprop.data.utils import get_data 10 | from chemprop.utils import load_checkpoint 11 | 12 | 13 | def visualize_attention(args: Namespace): 14 | """Visualizes attention weights.""" 15 | print('Loading data') 16 | data = get_data(path=args.data_path) 17 | smiles = data.smiles() 18 | print(f'Data size = {len(smiles):,}') 19 | 20 | print(f'Loading model from "{args.checkpoint_path}"') 21 | model = load_checkpoint(args.checkpoint_path, cuda=args.cuda) 22 | mpn = model[0] 23 | 24 | for i in trange(0, len(smiles), args.batch_size): 25 | smiles_batch = smiles[i:i + args.batch_size] 26 | mpn.viz_attention(smiles_batch, viz_dir=args.viz_dir) 27 | 28 | 29 | if __name__ == '__main__': 30 | parser = ArgumentParser() 31 | parser.add_argument('--data_path', type=str, required=True, 32 | help='Path to data CSV file') 33 | parser.add_argument('--viz_dir', type=str, required=True, 34 | help='Path where attention PNGs will be saved') 35 | parser.add_argument('--checkpoint_path', type=str, required=True, 36 | help='Path to a model checkpoint') 37 | parser.add_argument('--batch_size', type=int, default=50, 38 | help='Batch size') 39 | parser.add_argument('--no_cuda', action='store_true', default=False, 40 | help='Turn off cuda') 41 | args = parser.parse_args() 42 | 43 | # Cuda 44 | args.cuda = not args.no_cuda and torch.cuda.is_available() 45 | del args.no_cuda 46 | 47 | # Create directory for preds path 48 | os.makedirs(args.viz_dir, exist_ok=True) 49 | 50 | visualize_attention(args) 51 | -------------------------------------------------------------------------------- /setup.cfg: -------------------------------------------------------------------------------- 1 | [metadata] 2 | description-file = README.md 3 | -------------------------------------------------------------------------------- /setup.py: -------------------------------------------------------------------------------- 1 | from setuptools import find_packages, setup 2 | 3 | with open('requirements.txt') as f: 4 | requirements = [line.strip() for line in f if not line.startswith('git+http')] 5 | 6 | setup( 7 | name='chemprop', 8 | version='1.0', 9 | author='Wengong Jin, Kyle Swanson, Kevin Yang', 10 | author_email='wengong@csail.mit.edu, swansonk@mit.edu, yangk@mit.edu', 11 | description='Molecular Property Prediction with Message Passing Neural Networks', 12 | url='https://github.com/wengong-jin/chemprop', 13 | license='MIT', 14 | packages=find_packages(), 15 | install_requires=requirements, 16 | keywords=['chemistry', 'machine learning', 'property prediction', 'message passing neural network'], 17 | ) 18 | -------------------------------------------------------------------------------- /static/jsme/.directory: -------------------------------------------------------------------------------- 1 | [Dolphin] 2 | SortOrder=1 3 | Timestamp=2015,5,31,16,7,38 4 | Version=3 5 | ViewMode=1 6 | -------------------------------------------------------------------------------- /static/jsme/JME_to_JSME.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | JME to JSME migration. Part 1 4 | 5 | 6 | 7 | 8 |

JME to JSME migration

9 |
10 | You have to enable JavaScript in your browser to use JSME ! 11 |
12 | The variable name "jme" refers to the applet document.jme . 13 | 14 |
15 | The applet can also be accessed via document.jsapplets[0]. 16 | 17 |
18 | 19 | JSME in depict mode: 20 |
21 | You have to enable JavaScript on your machine ! 22 | 23 | 24 | 25 | You have to enable JavaScript in your browser to use JSME ! 26 | 27 | 28 |
29 | 30 | 31 | -------------------------------------------------------------------------------- /static/jsme/JME_to_JSME_simple.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | JME Example 6 | 7 | 8 | 9 |
10 | You have to enable JavaScript in your browser to use JSME ! 11 |
12 | 13 | -------------------------------------------------------------------------------- /static/jsme/JSME_SVG.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 28 | 29 | 30 | 31 |
32 | 33 | 34 |
35 | 36 | 38 | 39 | 40 | 41 | -------------------------------------------------------------------------------- /static/jsme/JSME_atom_highlight_demo.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 42 | 43 | 44 | 45 | 46 | 47 | 48 |

JSME atom highlight

49 | 50 | 51 | 52 | 53 | 54 |
55 | 56 | 57 | 58 | 59 | 60 | 61 | 62 | 63 | 64 | 65 | 66 | 70 | 71 | 72 | 73 | 76 | 77 | 78 | 79 | 80 | 81 | 82 | 83 | 88 | 89 | 90 |
67 | 68 | 69 |
74 | Atom highlighted by mouse over: 75 |
84 | Atom colors: 85 | 86 | 87 |
91 | 92 | 93 |

Atom background color

94 | There are 6 predefined pastel colors available: numbered 1 to 6. The color specification is character string containing a comma separated list of atom number and color index.
95 | Example: 1,2 : color atom #1 with color # 2
96 | Example: 1,2,10,4 : color atom #1 with color # 2 and color atom #10 with color #4
97 | 98 |

Bond background color

99 | API not implemented yet. 100 | 101 | 102 |

API

103 |

104 | Note: All atom and molecule indices start at 1. 105 |

106 | 107 |

108 | public void setAtomToHighLight(int molIndex, int atomIndex)
109 | The hightlight is temporary. 110 |

111 | 112 |

113 | public void setNotifyAtomHighLightChangeJSfunction(String notifyAtomHighLightJSfunction)
114 | Receive a notification when JSME detects a mouse over one atom. The argument is the name of a JavaScript function that receives two arguments: the molecule index and the atom index.
115 | To cancel, set the argument to null. 116 |

117 | 118 | 119 |

120 | public void setAtomBackgroundColors(int molIndex, String atomAndColorCSV)
121 | atomAndColorCSV must be an integer between 0 and 6. 0 means no background color. 122 |

123 | 124 |

125 | public void resetAtomColors(int molIndex) 126 |

127 | 128 | 129 | 130 | 131 | 132 | -------------------------------------------------------------------------------- /static/jsme/JSME_autoresize.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 15 | 16 | 17 | 18 |

Resize the window and the applet should resize accordingly. Does not work with IE678 and Opera 12.

19 |
20 | 21 | 22 | 23 | 24 | -------------------------------------------------------------------------------- /static/jsme/JSME_depict.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 25 | 26 | 27 | 28 |

Depict mode with setNotifyAtomHighLightChangeJSfunction

29 |
30 |
31 | Atom highlighted by mouse over: 32 |
33 |
34 |
35 | 36 | 100 | 101 |
102 | 103 | 104 | 105 | 106 | -------------------------------------------------------------------------------- /static/jsme/JSME_depict_action.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 28 | 29 | 30 | 31 |

Depict mode with setNotifyAtomHighLightChangeJSfunction and the star option enabled

32 |
33 |
34 | Atom highlighted by mouse over: 35 |
36 |
37 | 38 | 39 | 40 | 41 | -------------------------------------------------------------------------------- /static/jsme/JSME_depict_action_callback.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 53 | 54 | 55 | 56 |

Depict mode with the callback "AtomHighLight" and the star option enabled

57 |
58 |
59 | Atom highlighted by mouse over: 60 | Bond highlighted by mouse over: 61 |
62 | Atom clicked: 63 | Bond clicked: 64 |
65 |
66 | Event: 67 | 68 | 69 | 70 | 71 | -------------------------------------------------------------------------------- /static/jsme/JSME_depict_edit_toggle.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 52 | 53 | 54 | 55 | 56 |

Toggle depict - edit

57 | This example shows how to switch back and forth between depict and edit mode. 58 |
59 |
60 | 61 | Draw a chemical structure.
62 | Please click on the area below to edit the structure. 63 | 64 |
65 |
66 | 67 | 68 | 69 | 70 | 71 | 72 | 73 | 74 | -------------------------------------------------------------------------------- /static/jsme/JSME_depict_smiles.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 |

8 | JSME in depict mode initialized with a SMILES 9 |

10 | 11 | SMILES conversion and 2D coordinates computation are performed by the Open Source Chemistry Library OpenChemLib 12 | 13 | 38 | 39 |

Taxol

40 | Width changes when the browser window changes, the molecule is rescaled to fit if needed 41 | 42 |
43 | 44 | 45 | 46 |
47 | 48 | 49 | 57 | 58 | 59 | 60 | -------------------------------------------------------------------------------- /static/jsme/JSME_dnd_demo.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 20 | 21 | 22 |

JSME drag and drop

23 | If drag and drop is supported by your browser, an icon will be displayed at the bottom right of the editor. 24 | 25 | 26 | 27 | 28 |
29 |
30 | 31 |

Testing the DnD capabilities of JSME:

32 |
    33 |
  1. Draw a structure on the top editor
    2. 34 |
  2. Drag it out and drop it on the surface of the bottom editor
    3. 35 |
  3. Drag it out and drop it on your computer local file system to save it (does not work in Windows)
    4. 36 |
  4. Drag it out and drop it on the window of another program, e.g. text editor
    5. 37 |
  5. Drag a MOL or SDF file from your computer local file system and drop it on to the bottom or top editor 38 |
39 | 40 | Does not work yet on iPad 4 and Android devices. 41 | 42 | 43 | 44 | 45 | 46 | 47 | -------------------------------------------------------------------------------- /static/jsme/JSME_editor_plus_SVG.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 26 | 27 | 28 | 29 |
30 | 31 | 32 |
33 | 34 | 36 | 37 | 38 | 39 | -------------------------------------------------------------------------------- /static/jsme/JSME_hidden.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 24 | 25 | 26 | 27 | 28 |

JSME in a hidden div

29 | 30 |

The size is fixed (defined in px)

31 | 32 | 33 | 34 |
35 |
36 |
37 |
38 | 39 |
40 |
41 |
42 |

The size is dynamic (defined in %)

43 | Note: jsmeApplet2.repaint() must be called after displaying. 44 |
45 | 46 | 47 | 48 |
49 |
50 |
51 |
52 | 53 | 54 | 55 | -------------------------------------------------------------------------------- /static/jsme/JSME_minimal.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 12 | 13 | 14 | 15 |
16 | 17 | 18 | 19 | 20 | -------------------------------------------------------------------------------- /static/jsme/JSME_minimal2.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 13 | 14 | 15 | 16 |
17 | 18 | 19 | 20 | 21 | 22 | -------------------------------------------------------------------------------- /static/jsme/JSME_minimal3.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 14 | 15 | 16 | 17 |
18 | 19 | 20 | 21 | 22 | 23 | -------------------------------------------------------------------------------- /static/jsme/JSME_resize.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 12 | 13 | 14 | 15 | 16 |
17 | 18 | 19 | 20 | 27 | 28 | 29 | 30 | -------------------------------------------------------------------------------- /static/jsme/JSME_smiles_atom_highlight.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | SMILES Hilight demo 5 | 6 | 9 | 10 | 36 | 37 | 38 | 39 |

SMILES Highlight demo

40 |

41 | Draw a structure, click on the star icon of the editor and click on one or more atoms 42 |

43 |
44 |
45 | Canonical SMILES 46 |
47 |
48 | SMILES: 49 | 50 | 51 | 52 | 53 | -------------------------------------------------------------------------------- /static/jsme/JSME_template.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | JME template creation 5 | 6 | 7 | 8 | 9 | 24 |

This page shows how to create custom templates for JSME

25 | 26 | To create a template, one has to: 27 |
    28 |
  1. draw a chemical structure
    2. 29 |
  2. mark the connection atom using the blue circle
    3. 30 |
  3. save the template as JME string (button "get JME string")
    4. 31 |
32 |

33 | The atom that should act as a connection point (root) of the template should by marked by :1 (for example the oxygen serving as root should have atomic symbol O:1). 34 |

35 |

36 | 37 | You have to enable JavaScript in your browser to use JSME!
38 | 39 | 40 | 41 | 42 | 43 | 44 |
45 |
46 | 47 | To use a predefined template: 48 |
    49 |
  1. draw a chemical structure
    2. 50 |
  2. input a name
    3. 51 |
  3. Click on "set template"
    4. 52 |
  4. Click on one atom of the chemical structure to attach the template, or alternatively use the NEW button and click elsewhere in the drawing area
    5. 53 |
54 |
55 | Template name:
56 | 57 |

58 |

59 | See this page for real example. 60 | 61 | 62 | -------------------------------------------------------------------------------- /static/jsme/api_javadoc/allclasses-frame.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | All Classes 7 | 8 | 9 | 10 | 11 | 12 |

All Classes

13 |
14 | 18 |
19 | 20 | 21 | -------------------------------------------------------------------------------- /static/jsme/api_javadoc/allclasses-noframe.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | All Classes 7 | 8 | 9 | 10 | 11 | 12 |

All Classes

13 |
14 | 18 |
19 | 20 | 21 | -------------------------------------------------------------------------------- /static/jsme/api_javadoc/constant-values.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | Constant Field Values 7 | 8 | 9 | 10 | 11 | 12 | 22 | 25 | 26 |
27 | 28 | 29 |
Skip navigation links
30 | 31 | 32 | 33 | 41 |
42 |
43 | 47 | 51 | 54 |
55 | 65 |
66 | 67 | 68 |
69 | 70 |
71 |

Constant Field Values

72 |

Contents

73 |
74 | 75 |
76 | 77 | 78 |
Skip navigation links
79 | 80 | 81 | 82 | 90 |
91 |
92 | 96 | 100 | 103 |
104 | 114 |
115 | 116 | 117 |
118 | 119 | 120 | 121 | -------------------------------------------------------------------------------- /static/jsme/api_javadoc/export/client/package-frame.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | export.client 7 | 8 | 9 | 10 | 11 | 12 |

export.client

13 |
14 |

Classes

15 | 19 |
20 | 21 | 22 | -------------------------------------------------------------------------------- /static/jsme/api_javadoc/export/client/package-tree.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | export.client Class Hierarchy 7 | 8 | 9 | 10 | 11 | 12 | 22 | 25 | 26 |
27 | 28 | 29 |
Skip navigation links
30 | 31 | 32 | 33 | 41 |
42 |
43 | 47 | 51 | 54 |
55 | 65 |
66 | 67 | 68 |
69 | 70 |
71 |

Hierarchy For Package export.client

72 |
73 |
74 |

Class Hierarchy

75 | 83 |
84 | 85 |
86 | 87 | 88 |
Skip navigation links
89 | 90 | 91 | 92 | 100 |
101 |
102 | 106 | 110 | 113 |
114 | 124 |
125 | 126 | 127 |
128 | 129 | 130 | 131 | -------------------------------------------------------------------------------- /static/jsme/api_javadoc/index.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | Generated Documentation (Untitled) 7 | 59 | 60 | 61 | 62 | 63 | 64 | <noscript> 65 | <div>JavaScript is disabled on your browser.</div> 66 | </noscript> 67 | <h2>Frame Alert</h2> 68 | <p>This document is designed to be viewed using the frames feature. If you see this message, you are using a non-frame-capable web client. Link to <a href="export/client/package-summary.html">Non-frame version</a>.</p> 69 | 70 | 71 | 72 | -------------------------------------------------------------------------------- /static/jsme/api_javadoc/package-list: -------------------------------------------------------------------------------- 1 | export.client 2 | -------------------------------------------------------------------------------- /static/jsme/api_javadoc/script.js: -------------------------------------------------------------------------------- 1 | function show(type) 2 | { 3 | count = 0; 4 | for (var key in methods) { 5 | var row = document.getElementById(key); 6 | if ((methods[key] & type) != 0) { 7 | row.style.display = ''; 8 | row.className = (count++ % 2) ? rowColor : altColor; 9 | } 10 | else 11 | row.style.display = 'none'; 12 | } 13 | updateTabs(type); 14 | } 15 | 16 | function updateTabs(type) 17 | { 18 | for (var value in tabs) { 19 | var sNode = document.getElementById(tabs[value][0]); 20 | var spanNode = sNode.firstChild; 21 | if (value == type) { 22 | sNode.className = activeTableTab; 23 | spanNode.innerHTML = tabs[value][1]; 24 | } 25 | else { 26 | sNode.className = tableTab; 27 | spanNode.innerHTML = "" + tabs[value][1] + ""; 28 | } 29 | } 30 | } 31 | -------------------------------------------------------------------------------- /static/jsme/bootstrap.css: -------------------------------------------------------------------------------- 1 | /*! 2 | * Bootstrap v3.3.2 (http://getbootstrap.com) 3 | * Copyright 2011-2015 Twitter, Inc. 4 | * Licensed under MIT (https://github.com/twbs/bootstrap/blob/master/LICENSE) 5 | *//*! normalize.css v3.0.2 | MIT License | git.io/normalize */ 6 | 7 | 8 | 9 | img { 10 | vertical-align: middle; 11 | } 12 | 13 | -------------------------------------------------------------------------------- /static/jsme/bootstrap.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | Structure modified and atom mark demo 5 | 6 | 7 | 8 | 11 | 12 | 35 | 36 | 37 | 38 |

Structure modified and atom mark demo

39 |

40 | Draw a structure, click on the star icon of the editor and click on one or more atoms 41 |

42 |
43 | 44 |
45 | 46 | 47 | 48 | 49 | -------------------------------------------------------------------------------- /static/jsme/doc.css: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | div.example_code { 5 | background-color: #F6F4F0; 6 | border: 1px solid #D4D4D4; 7 | font-family: "courier new"; 8 | font-size: 90%; 9 | padding: 5px; 10 | width: 98%; 11 | border-style:solid; 12 | border-color:white; 13 | border-width:15px; 14 | width: 90%; 15 | } 16 | div.example_code_jsme { 17 | background-color: #F6F4F8; 18 | border: 1px solid #D4D4D4; 19 | font-family: "courier new"; 20 | font-size: 90%; 21 | padding: 5px; 22 | border-style:solid; 23 | border-color:white; 24 | border-width:15px; 25 | width: 90%; 26 | } 27 | 28 | -------------------------------------------------------------------------------- /static/jsme/index.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | JSME Molecule Editor 6 | 7 | 8 |

JSME Molecule Editor Distribution

9 | 10 | This distribution contains 3 directories:
11 | 22 |

23 | Files contained in the current directory should help you to implement JSME easily:

24 | 25 | doc.html provides basic information about JSME implementation into web page - start with this document
26 |

27 | 28 |

29 | Example pages:
30 |

47 |

48 | And some other files:
49 |

53 |

54 | If you are using the JSME cite please the following article:
55 | B. Bienfait and P. Ertl, JSME: a free molecule editor in JavaScript, J. Cheminformatics 5:24 (2013) 56 |

57 | Happy molecule editing! 58 | 59 | -------------------------------------------------------------------------------- /static/jsme/jme_examples/jme_example2.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | JME Example 2 4 | 5 | 33 | 34 | 35 | 36 |

Example of Structure Input

37 | 38 |
39 | Press the Draw Molecule button, draw molecule and then press Submit Molecule button in the editor window.

40 | 41 |    42 | 43 |    44 | 45 |

46 | 47 |

48 |

49 | 50 | 51 | 52 | -------------------------------------------------------------------------------- /static/jsme/jme_examples/jme_example4.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 5 | 6 | 7 | 8 | JME Example 4 9 | 47 | 48 | 49 | 50 | 52 |

53 |

Adding additional functionality to JME by JavaScript

54 | Capabilities of JME may be enhanced by adding relatively simple 55 | JavaScript code to the page containing JME. 56 |
57 |

Substituent menu

58 | This option allows fast and convenient creation of common organic 59 | substituents. Check the JavaScript code in this page to see how the 60 | substituent menu should be implemented. 61 |

62 |

Molecule persistency

63 | Several users asked for a "persistent molecule" in the JME Editor. That 64 | means that when they visit another page and then come back, the JME 65 | should remember the molecule recently edited. It is relatively easy to 66 | implement this feature. It requires just to add two JavaScript 67 | functions to the page (saveJMECookie and read JMECookie), as shown in 68 | the source of this page. Additionally, respective event handlers must 69 | be added into the page BODY tag. 70 |
Whenever the HTML page with the editor is abandoned, the 71 | actual molecule is stored into a so called "cookie" (a small piece of 72 | persistent information associated with this particular web page) and 73 | when the page is revisited (even after several weeks), the molecule is 74 | automatically restored. 75 |

76 |

77 | 78 | 79 | 80 | 96 | 97 |
 
98 |
You have to enable Java in your 100 | browser.
101 | 103 | 104 | 106 | 110 | 111 |
JME 105 | Editor courtesy of Peter Ertl, NovartisJME help
112 |
113 | 114 | -------------------------------------------------------------------------------- /static/jsme/jme_examples/jme_window.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | JME Molecular Editor 4 | 6 | 7 | 8 | 9 | 28 | 29 | 30 | 31 |
32 | 33 |
You have to enable Java and JavaScritpt on your 35 | machine !
36 |
37 |       41 |
42 |
43 | 44 | -------------------------------------------------------------------------------- /static/jsme/jsme/396F806CD63ABD414BFBB9D57429F05B.cache.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/static/jsme/jsme/396F806CD63ABD414BFBB9D57429F05B.cache.png -------------------------------------------------------------------------------- /static/jsme/jsme/40BAF81124143A595056A9CCA0E9DBBA.cache.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/static/jsme/jsme/40BAF81124143A595056A9CCA0E9DBBA.cache.png -------------------------------------------------------------------------------- /static/jsme/jsme/4841BDE9DC293BA35F7762B4D8EFD236.cache.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/static/jsme/jsme/4841BDE9DC293BA35F7762B4D8EFD236.cache.png -------------------------------------------------------------------------------- /static/jsme/jsme/A6DBDE07E3A8F66E8959A4F32505E16B.cache.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/static/jsme/jsme/A6DBDE07E3A8F66E8959A4F32505E16B.cache.png -------------------------------------------------------------------------------- /static/jsme/jsme/DF7764EEC1903CD03C9545B354D8D8E4.cache.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/static/jsme/jsme/DF7764EEC1903CD03C9545B354D8D8E4.cache.png -------------------------------------------------------------------------------- /static/jsme/jsme/clear.cache.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wengong-jin/chemprop/3ad3577367d8a53f28aade0be41b56b1f25b6125/static/jsme/jsme/clear.cache.gif -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/222ADBFEC322C2723C6ED2C4FB31B217/3.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback3('w(685,679,bv);_.Ad=function(){this.a.j&&W2(this.a.j);this.a.j=new a3(0,this.a)};B(sW)(3);\n//@ sourceURL=3.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/222ADBFEC322C2723C6ED2C4FB31B217/4.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback4('w(687,679,bv);_.Ad=function(){this.a.pc&&W2(this.a.pc);this.a.pc=new a3(1,this.a)};B(sW)(4);\n//@ sourceURL=4.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/222ADBFEC322C2723C6ED2C4FB31B217/5.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback5('w(688,679,bv);_.Ad=function(){this.a.y&&(W2(this.a.y),this.a.y=null);0==this.a.ob.v&&(this.a.y=new a3(2,this.a))};B(sW)(5);\n//@ sourceURL=5.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/222ADBFEC322C2723C6ED2C4FB31B217/7.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback7('var gma="New atom map",hma="atom map";function j$(a){a=a.Jc.b.c++;return Xo+a}w(475,476,xW);_.Ne=function(){return j$(this)};_.Ye=function(){this.Tc=new k$((KJ(),m),this)};_.cf=function(){return(null==this.Qc&&(this.Qc=j$(this)),this.Qc)+Jd+this.Tc.We()+Jd+this.Tc.Xe()+Jd+this.Yc+eu+this.Nc+(this.Xc?m:",hidden")+",title="+this.q+(this.p?",resizable":m)+(this.Tc.c.C?",modal":m)};\nfunction l$(){var a,b,c;b=new O1(QW);a=new O1(IW);this.a=new DL(Xe,4);new W1;this.c=new m$(this,b);this.c.Tc.c.db=!0;XJ(this.c,new cK);x1(this.c,new c2(gma),null);x1(this.c,b,null);x1(this.c,a,null);XJ(this.c,new L1(2));c=new bK;x1(c,new d2(hma,1),null);x1(c,this.a,null);x1(this.c,c,lW);c=new bK;x1(c,b,null);x1(c,a,null);x1(this.c,c,kW);rK(this.c.r,new n$(this))}w(663,1,{},l$);_.a=null;_.b=null;_.c=null;\nfunction m$(a,b){this.a=a;this.b=b;V1.call(this);this.q=m;this.Tc&&this.Tc.Lf(m);this.Tc.c.db=!1;U1(this,!0)}w(664,475,xW,m$);_.Me=function(a){var b;S1(this);if(eE(a.f)===eE(this.b))try{b=pM(Wz(this.a.a.Tc.a.pb,Yt),10);var c=this.a.b,d=c.c,e=c.a;hR(d,c.e).b[e].f=b;MN(d,"changeAtomMap",d.o,e,0);uQ(d)}catch(f){if(f=Az(f),H(f,102))p2(this.a.b.c,"invalid atom map");else throw f;}return!0};_.a=null;_.b=null;function n$(a){this.a=a}w(665,521,AW,n$);_.Df=function(){S1(this.a.c)};_.a=null;w(678,679,bv);\n_.Ad=function(){if(this.b.a==(hN(),Wg)){!this.a.b&&(this.a.b=new l$);var a=this.a.b,b=this.a,c;if(0>1;f=~~(pr()-cr(a.pb,ej))>>1;OY(a,fn(lr($doc.body)+e,0),fn(($doc.body.scrollTop||0)+f,0));d||((a.ab=c)?(a.pb.style[ah]=Bj,a.pb.style[Kk]=Lk,Em(a.gb,200)):a.pb.style[Kk]=Lk)}finally{a.w=!0}}function X4(a){a.i=(new SW(a.j)).Tc.Nf();Fy(a.i,new Y4(a),(Et(),Et(),Ft));a.d=z(GB,u,51,[a.i])}\nfunction Z4(){BY();var a,b,c,d,e;$Y.call(this,(sZ(),tZ),null,!0);this.ei();this.db=!0;a=new S4(this.k);this.f=new T4;this.f.pb.style[Nk]=Hc;ry(this.f,Hc);this.ci();sY(this,"400px");e=new V4;e.pb.style[di]=Hc;e.e[Tg]=10;c=(aA(),bA);e.a=c;U4(e,a);U4(e,this.f);this.e=new pA;this.e.e[Tg]=20;for(b=this.d,c=0,d=b.length;cb)throw new tL("Length must be a positive integer. Length: "+b);if(b>dr(a.pb,Ik).length)throw new tL("From Index: 0 To Index: "+b+" Text Length: "+dr(a.pb,Ik).length);try{a.pb.setSelectionRange(0,0+b)}catch(c){}}};_.a=null;function a5(a){X4(a);a.a=(new SW(a.b)).Tc.Nf();Fy(a.a,new b5(a),(Et(),Et(),Ft));a.d=z(GB,u,51,[a.a,a.i])}\nfunction c5(a){a.j=wQ;a.k="Paste the text to import into the text area below.";a.b="Accept";rZ(a.I.b,"Paste")}function d5(a){BY();Z4.call(this);this.c=a}w(790,784,gQ,d5);_.ci=function(){a5(this)};_.di=function(){ry(this.f,"150px")};_.ei=function(){c5(this)};_.Me=function(){ZY(this);Tq((Qq(),Rq),new e5(this))};_.a=null;_.b=null;_.c=null;function f5(a){BY();d5.call(this,a)}w(789,790,gQ,f5);_.ci=function(){var a;a5(this);a=new R4;Fy(a,new g5(this),(OU(),OU(),PU));this.d=z(GB,u,51,[this.a,a,this.i])};\n_.di=function(){ry(this.f,"150px");EF(new h5(this),this.f)};_.ei=function(){c5(this);this.k+=" Or drag and drop a file on it."};function g5(a){this.a=a}w(791,1,{},g5);_.Md=function(a){var b,c;b=new FileReader;a=(c=nr(a.a),c.files[0]);i5(b,new j5(this));b.readAsText(a)};_.a=null;function j5(a){this.a=a}w(792,1,{},j5);_.ag=function(a){LE();rB(this.a.a.f,a)};_.a=null;function h5(a){this.a=a;this.b=new k5(this);this.c=this.d=1}w(793,568,{},h5);_.a=null;function k5(a){this.a=a}w(794,1,{},k5);\n_.ag=function(a){this.a.a.f.pb[Ik]=null!=a?a:m};_.a=null;function b5(a){this.a=a}w(798,1,{},b5);_.Nd=function(){if(this.a.c){var a=this.a.c,b;b=new GE(a.a,0,dr(this.a.f.pb,Ik));MF(a.a.a,b.a)}IY(this.a,!1)};_.a=null;function e5(a){this.a=a}w(799,1,{},e5);_.pd=function(){Ay(this.a.f.pb,!0);Qz.Qe(this.a.f.pb)};_.a=null;w(800,1,am);_.Ed=function(){var a,b;a=new l5(this.a);void 0!=$wnd.FileReader?b=new f5(a):b=new d5(a);uY(b);W4(b)};function l5(a){this.a=a}w(801,1,{},l5);_.a=null;w(802,1,am);\n_.Ed=function(){var a;a=new Z4;var b=this.a,c,d;rB(a.f,b);c=(d=UL(b,"\\r\\n|\\r|\\n|\\n\\r"),d.length);1>=c&&(c=~~(b.length/16));ry(a.f,20*(10>c+1?c+1:10)+uj);Tq((Qq(),Rq),new $4(a));uY(a);W4(a)};function i5(a,b){a.onload=function(a){b.ag(a.target.result)}}X(784);X(790);X(789);X(801);X(787);X(788);X(798);X(799);X(791);X(792);X(793);X(794);X(408);X(457);X(448);X(404);C(fQ)(8);\n//@ sourceURL=8.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/6ABB8447ACAB1353A478923AC9C0550B/3.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback3('w(680,674,Hl);_.Ad=function(){this.a.j&&sW(this.a.j);this.a.j=new xW(0,this.a)};B(qO)(3);\n//@ sourceURL=3.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/6ABB8447ACAB1353A478923AC9C0550B/4.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback4('w(682,674,Hl);_.Ad=function(){this.a.pc&&sW(this.a.pc);this.a.pc=new xW(1,this.a)};B(qO)(4);\n//@ sourceURL=4.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/6ABB8447ACAB1353A478923AC9C0550B/5.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback5('w(683,674,Hl);_.Ad=function(){this.a.y&&(sW(this.a.y),this.a.y=null);0==this.a.ob.v&&(this.a.y=new xW(2,this.a))};B(qO)(5);\n//@ sourceURL=5.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/6ABB8447ACAB1353A478923AC9C0550B/7.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback7('function h3(a){a=a.Jc.b.c++;return dh+a}w(470,471,vO);_.Je=function(){return h3(this)};_.Ue=function(){this.Tc=new i3((DA(),m),this)};_.$e=function(){return(null==this.Qc&&(this.Qc=h3(this)),this.Qc)+Pa+this.Tc.Se()+Pa+this.Tc.Te()+Pa+this.Yc+Ak+this.Nc+(this.Xc?m:",hidden")+",title="+this.q+(this.p?",resizable":m)+(this.Tc.c.C?",modal":m)};\nfunction j3(){var a,b,c;b=new hV(PO);a=new hV(HO);this.a=new CC(bc,4);new pV;this.c=new k3(this,b);this.c.Tc.c.db=!0;RA(this.c,new XA);QU(this.c,new xV("New atom map"),null);QU(this.c,b,null);QU(this.c,a,null);RA(this.c,new dV(2));c=new WA;QU(c,new yV("atom map",1),null);QU(c,this.a,null);QU(this.c,c,jO);c=new WA;QU(c,b,null);QU(c,a,null);QU(this.c,c,iO);lB(this.c.r,new l3(this))}w(658,1,{},j3);_.a=null;_.b=null;_.c=null;\nfunction k3(a,b){this.a=a;this.b=b;oV.call(this);this.q=m;this.Tc&&this.Tc.Hf(m);this.Tc.c.db=!1;nV(this,!0)}w(659,470,vO,k3);_.Ie=function(a){var b;lV(this);if(Tu(a.f)===Tu(this.b))try{b=qD(Oq(this.a.a.Tc.a.pb,sk),10);var c=this.a.b,d=c.c,e=c.a;nI(d,c.e).b[e].f=b;RE(d,"changeAtomMap",d.o,e,0);AH(d)}catch(f){if(f=pq(f),H(f,102))KV(this.a.b.c,"invalid atom map");else throw f;}return!0};_.a=null;_.b=null;function l3(a){this.a=a}w(660,516,yO,l3);_.zf=function(){lV(this.a.c)};_.a=null;w(673,674,Hl);\n_.Ad=function(){if(this.b.a==(kE(),fe)){!this.a.b&&(this.a.b=new j3);var a=this.a.b,b=this.a,c;if(0>1;f=vr()-gr(a.pb,ej)>>1;iZ(a,gn(xr($doc)+e,0),gn(yr($doc)+f,0));d||((a.ab=c)?(zB(a.pb,Bj),a.pb.style[Ik]=Jk,Dm(a.gb,200)):a.pb.style[Ik]=Jk)}finally{a.w=!0}}function r5(a){a.i=(new lX(a.j)).Tc.Mf();Zy(a.i,new s5(a),(Xt(),Xt(),Yt));a.d=z(YB,q,52,[a.i])}\nfunction t5(){WY();var a,b,c,d,e;uZ.call(this,(NZ(),OZ),null,!0);this.di();this.db=!0;a=new m5(this.k);this.f=new n5;this.f.pb.style[Lk]=Gc;Ly(this.f,Gc);this.bi();NY(this,"400px");e=new p5;e.pb.style[ci]=Gc;e.e[Rg]=10;c=(tA(),uA);e.a=c;o5(e,a);o5(e,this.f);this.e=new IA;this.e.e[Rg]=20;for(b=this.d,c=0,d=b.length;cb)throw new IL("Length must be a positive integer. Length: "+b);if(b>hr(a.pb,Gk).length)throw new IL("From Index: 0 To Index: "+b+" Text Length: "+hr(a.pb,Gk).length);try{a.pb.setSelectionRange(0,0+b)}catch(c){}}};_.a=null;function v5(a){r5(a);a.a=(new lX(a.b)).Tc.Mf();Zy(a.a,new w5(a),(Xt(),Xt(),Yt));a.d=z(YB,q,52,[a.a,a.i])}\nfunction x5(a){a.j=NQ;a.k="Paste the text to import into the text area below.";a.b="Accept";MZ(a.I.b,"Paste")}function y5(a){WY();t5.call(this);this.c=a}w(794,788,zQ,y5);_.bi=function(){v5(this)};_.ci=function(){Ly(this.f,"150px")};_.di=function(){x5(this)};_.Me=function(){tZ(this);Sq((Pq(),Qq),new z5(this))};_.a=null;_.b=null;_.c=null;function A5(a){WY();y5.call(this,a)}w(793,794,zQ,A5);_.bi=function(){var a;v5(this);a=new l5;Zy(a,new B5(this),(gV(),gV(),hV));this.d=z(YB,q,52,[this.a,a,this.i])};\n_.ci=function(){Ly(this.f,"150px");SF(new C5(this),this.f)};_.di=function(){x5(this);this.k+=" Or drag and drop a file on it."};function B5(a){this.a=a}w(795,1,{},B5);_.Md=function(a){var b,c;b=new FileReader;a=(c=a.a.target,c.files[0]);D5(b,new E5(this));b.readAsText(a)};_.a=null;function E5(a){this.a=a}w(796,1,{},E5);_._f=function(a){ZE();KB(this.a.a.f,a)};_.a=null;function C5(a){this.a=a;this.b=new F5(this);this.c=this.d=1}w(797,572,{},C5);_.a=null;function F5(a){this.a=a}w(798,1,{},F5);\n_._f=function(a){this.a.a.f.pb[Gk]=null!=a?a:m};_.a=null;function w5(a){this.a=a}w(802,1,{},w5);_.Nd=function(){if(this.a.c){var a=this.a.c,b;b=new UE(a.a,0,hr(this.a.f.pb,Gk));$F(a.a.a,b.a)}cZ(this.a,!1)};_.a=null;function z5(a){this.a=a}w(803,1,{},z5);_.pd=function(){Uy(this.a.f.pb,!0);this.a.f.pb.focus()};_.a=null;w(804,1,Yl);_.Ed=function(){var a,b;a=new G5(this.a);void 0!=$wnd.FileReader?b=new A5(a):b=new y5(a);PY(b);q5(b)};function G5(a){this.a=a}w(805,1,{},G5);_.a=null;w(806,1,Yl);\n_.Ed=function(){var a;a=new t5;var b=this.a,c,d;KB(a.f,b);c=(d=iM(b,"\\r\\n|\\r|\\n|\\n\\r"),d.length);1>=c&&(c=~~(b.length/16));Ly(a.f,20*(10>c+1?c+1:10)+uj);Sq((Pq(),Qq),new u5(a));PY(a);q5(a)};function D5(a,b){a.onload=function(a){b._f(a.target.result)}}X(788);X(794);X(793);X(805);X(791);X(792);X(802);X(803);X(795);X(796);X(797);X(798);X(411);X(460);X(451);X(407);C(vQ)(8);\n//@ sourceURL=8.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/C8A71BD2E1367E9BB43A1B9C25871BEE/3.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback3('w(713,707,Xl);_.Ed=function(){this.a.j&&vY(this.a.j);this.a.j=new AY(0,this.a)};C(uQ)(3);\n//@ sourceURL=3.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/C8A71BD2E1367E9BB43A1B9C25871BEE/4.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback4('w(715,707,Xl);_.Ed=function(){this.a.pc&&vY(this.a.pc);this.a.pc=new AY(1,this.a)};C(uQ)(4);\n//@ sourceURL=4.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/C8A71BD2E1367E9BB43A1B9C25871BEE/5.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback5('w(716,707,Xl);_.Ed=function(){this.a.y&&(vY(this.a.y),this.a.y=null);0==this.a.ob.v&&(this.a.y=new AY(2,this.a))};C(uQ)(5);\n//@ sourceURL=5.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/C8A71BD2E1367E9BB43A1B9C25871BEE/7.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback7('function e5(a){a=a.Jc.b.c++;return nh+a}w(497,498,vQ);_.Re=function(){return e5(this)};_.af=function(){this.Tc=new f5((yC(),m),this)};_.gf=function(){return(null==this.Qc&&(this.Qc=e5(this)),this.Qc)+nb+this.Tc.$e()+nb+this.Tc._e()+nb+this.Yc+Lk+this.Nc+(this.Xc?m:",hidden")+",title="+this.q+(this.p?",resizable":m)+(this.Tc.c.C?",modal":m)};\nfunction g5(){var a,b,c;b=new kX(UQ);a=new kX(MQ);this.a=new wE(xc,4);new sX;this.c=new h5(this,b);this.c.Tc.c.db=!0;MC(this.c,new SC);TW(this.c,new AX("New atom map"),null);TW(this.c,b,null);TW(this.c,a,null);MC(this.c,new gX(2));c=new RC;TW(c,new BX("atom map",1),null);TW(c,this.a,null);TW(this.c,c,mQ);c=new RC;TW(c,b,null);TW(c,a,null);TW(this.c,c,lQ);gD(this.c.r,new i5(this))}w(691,1,{},g5);_.a=null;_.b=null;_.c=null;\nfunction h5(a,b){this.a=a;this.b=b;rX.call(this);this.q=m;this.Tc&&this.Tc.Pf(m);this.Tc.c.db=!1;qX(this,!0)}w(692,497,vQ,h5);_.Qe=function(a){var b;oX(this);if(Ew(a.f)===Ew(this.b))try{b=mF(gr(this.a.a.Tc.a.pb,Fk),10);var c=this.a.b,d=c.c,e=c.a;pK(d,c.e).b[e].f=b;VG(d,"changeAtomMap",d.o,e,0);CJ(d)}catch(f){if(f=Gq(f),G(f,104))NX(this.a.b.c,"invalid atom map");else throw f;}return!0};_.a=null;_.b=null;function i5(a){this.a=a}w(693,543,BQ,i5);_.Hf=function(){oX(this.a.c)};_.a=null;w(706,707,Xl);\n_.Ed=function(){if(this.b.a==(oG(),Ae)){!this.a.b&&(this.a.b=new g5);var a=this.a.b,b=this.a,c;if(0>1;f=ur()-fr(a.pb,dj)>>1;hZ(a,fn(wr($doc)+e,0),fn(xr($doc)+f,0));d||((a.ab=c)?(yB(a.pb,Aj),a.pb.style[Hk]=Ik,Cm(a.gb,200)):a.pb.style[Hk]=Ik)}finally{a.w=!0}}function p5(a){a.i=(new kX(a.j)).Tc.Mf();Yy(a.i,new q5(a),(Wt(),Wt(),Xt));a.d=z(XB,q,52,[a.i])}\nfunction r5(){VY();var a,b,c,d,e;tZ.call(this,(MZ(),NZ),null,!0);this.di();this.db=!0;a=new k5(this.k);this.f=new l5;this.f.pb.style[Kk]=Gc;Ky(this.f,Gc);this.bi();MY(this,"400px");e=new n5;e.pb.style[ci]=Gc;e.e[Rg]=10;c=(sA(),tA);e.a=c;m5(e,a);m5(e,this.f);this.e=new HA;this.e.e[Rg]=20;for(b=this.d,c=0,d=b.length;cb)throw new HL("Length must be a positive integer. Length: "+b);if(b>gr(a.pb,Fk).length)throw new HL("From Index: 0 To Index: "+b+" Text Length: "+gr(a.pb,Fk).length);try{a.pb.setSelectionRange(0,0+b)}catch(c){}}};_.a=null;function t5(a){p5(a);a.a=(new kX(a.b)).Tc.Mf();Yy(a.a,new u5(a),(Wt(),Wt(),Xt));a.d=z(XB,q,52,[a.a,a.i])}\nfunction v5(a){a.j=MQ;a.k="Paste the text to import into the text area below.";a.b="Accept";LZ(a.I.b,"Paste")}function w5(a){VY();r5.call(this);this.c=a}w(794,788,yQ,w5);_.bi=function(){t5(this)};_.ci=function(){Ky(this.f,"150px")};_.di=function(){v5(this)};_.Me=function(){sZ(this);Rq((Oq(),Pq),new x5(this))};_.a=null;_.b=null;_.c=null;function y5(a){VY();w5.call(this,a)}w(793,794,yQ,y5);_.bi=function(){var a;t5(this);a=new j5;Yy(a,new z5(this),(fV(),fV(),gV));this.d=z(XB,q,52,[this.a,a,this.i])};\n_.ci=function(){Ky(this.f,"150px");RF(new A5(this),this.f)};_.di=function(){v5(this);this.k+=" Or drag and drop a file on it."};function z5(a){this.a=a}w(795,1,{},z5);_.Md=function(a){var b,c;b=new FileReader;a=(c=a.a.target,c.files[0]);B5(b,new C5(this));b.readAsText(a)};_.a=null;function C5(a){this.a=a}w(796,1,{},C5);_._f=function(a){YE();JB(this.a.a.f,a)};_.a=null;function A5(a){this.a=a;this.b=new D5(this);this.c=this.d=1}w(797,572,{},A5);_.a=null;function D5(a){this.a=a}w(798,1,{},D5);\n_._f=function(a){this.a.a.f.pb[Fk]=null!=a?a:m};_.a=null;function u5(a){this.a=a}w(802,1,{},u5);_.Nd=function(){if(this.a.c){var a=this.a.c,b;b=new TE(a.a,0,gr(this.a.f.pb,Fk));ZF(a.a.a,b.a)}bZ(this.a,!1)};_.a=null;function x5(a){this.a=a}w(803,1,{},x5);_.pd=function(){Ty(this.a.f.pb,!0);this.a.f.pb.focus()};_.a=null;w(804,1,Xl);_.Ed=function(){var a,b;a=new E5(this.a);void 0!=$wnd.FileReader?b=new y5(a):b=new w5(a);OY(b);o5(b)};function E5(a){this.a=a}w(805,1,{},E5);_.a=null;w(806,1,Xl);\n_.Ed=function(){var a;a=new r5;var b=this.a,c,d;JB(a.f,b);c=(d=hM(b,"\\r\\n|\\r|\\n|\\n\\r"),d.length);1>=c&&(c=~~(b.length/16));Ky(a.f,20*(10>c+1?c+1:10)+tj);Rq((Oq(),Pq),new s5(a));OY(a);o5(a)};function B5(a,b){a.onload=function(a){b._f(a.target.result)}}X(788);X(794);X(793);X(805);X(791);X(792);X(802);X(803);X(795);X(796);X(797);X(798);X(411);X(460);X(451);X(407);C(uQ)(8);\n//@ sourceURL=8.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/C9EEF554958AACEE6A060F620375E4FA/3.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback3('w(680,674,Hl);_.Ad=function(){this.a.j&&tW(this.a.j);this.a.j=new yW(0,this.a)};B(qO)(3);\n//@ sourceURL=3.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/C9EEF554958AACEE6A060F620375E4FA/4.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback4('w(682,674,Hl);_.Ad=function(){this.a.pc&&tW(this.a.pc);this.a.pc=new yW(1,this.a)};B(qO)(4);\n//@ sourceURL=4.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/C9EEF554958AACEE6A060F620375E4FA/5.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback5('w(683,674,Hl);_.Ad=function(){this.a.y&&(tW(this.a.y),this.a.y=null);0==this.a.ob.v&&(this.a.y=new yW(2,this.a))};B(qO)(5);\n//@ sourceURL=5.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/C9EEF554958AACEE6A060F620375E4FA/7.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback7('function j3(a){a=a.Jc.b.c++;return dh+a}w(470,471,vO);_.Je=function(){return j3(this)};_.Ue=function(){this.Tc=new k3((DA(),m),this)};_.$e=function(){return(null==this.Qc&&(this.Qc=j3(this)),this.Qc)+Pa+this.Tc.Se()+Pa+this.Tc.Te()+Pa+this.Yc+Ak+this.Nc+(this.Xc?m:",hidden")+",title="+this.q+(this.p?",resizable":m)+(this.Tc.c.C?",modal":m)};\nfunction l3(){var a,b,c;b=new iV(PO);a=new iV(HO);this.a=new CC(bc,4);new qV;this.c=new m3(this,b);this.c.Tc.c.db=!0;RA(this.c,new XA);RU(this.c,new yV("New atom map"),null);RU(this.c,b,null);RU(this.c,a,null);RA(this.c,new eV(2));c=new WA;RU(c,new zV("atom map",1),null);RU(c,this.a,null);RU(this.c,c,jO);c=new WA;RU(c,b,null);RU(c,a,null);RU(this.c,c,iO);lB(this.c.r,new n3(this))}w(658,1,{},l3);_.a=null;_.b=null;_.c=null;\nfunction m3(a,b){this.a=a;this.b=b;pV.call(this);this.q=m;this.Tc&&this.Tc.Hf(m);this.Tc.c.db=!1;oV(this,!0)}w(659,470,vO,m3);_.Ie=function(a){var b;mV(this);if(Tu(a.f)===Tu(this.b))try{b=qD(Oq(this.a.a.Tc.a.pb,sk),10);var c=this.a.b,d=c.c,e=c.a;nI(d,c.e).b[e].f=b;RE(d,"changeAtomMap",d.o,e,0);AH(d)}catch(f){if(f=pq(f),H(f,102))LV(this.a.b.c,"invalid atom map");else throw f;}return!0};_.a=null;_.b=null;function n3(a){this.a=a}w(660,516,yO,n3);_.zf=function(){mV(this.a.c)};_.a=null;w(673,674,Hl);\n_.Ad=function(){if(this.b.a==(kE(),fe)){!this.a.b&&(this.a.b=new l3);var a=this.a.b,b=this.a,c;if(0>1;f=~~(pr()-cr(a.pb,ej))>>1;PY(a,fn(lr($doc.body)+e,0),fn(($doc.body.scrollTop||0)+f,0));d||((a.ab=c)?(a.pb.style[ah]=Bj,a.pb.style[Kk]=Lk,Em(a.gb,200)):a.pb.style[Kk]=Lk)}finally{a.w=!0}}function Y4(a){a.i=(new TW(a.j)).Tc.Nf();Fy(a.i,new Z4(a),(Et(),Et(),Ft));a.d=z(GB,u,51,[a.i])}\nfunction $4(){CY();var a,b,c,d,e;aZ.call(this,(tZ(),uZ),null,!0);this.ei();this.db=!0;a=new T4(this.k);this.f=new U4;this.f.pb.style[Nk]=Hc;ry(this.f,Hc);this.ci();tY(this,"400px");e=new W4;e.pb.style[di]=Hc;e.e[Tg]=10;c=(aA(),bA);e.a=c;V4(e,a);V4(e,this.f);this.e=new pA;this.e.e[Tg]=20;for(b=this.d,c=0,d=b.length;cb)throw new tL("Length must be a positive integer. Length: "+b);if(b>dr(a.pb,Ik).length)throw new tL("From Index: 0 To Index: "+b+" Text Length: "+dr(a.pb,Ik).length);try{a.pb.setSelectionRange(0,0+b)}catch(c){}}};_.a=null;function b5(a){Y4(a);a.a=(new TW(a.b)).Tc.Nf();Fy(a.a,new c5(a),(Et(),Et(),Ft));a.d=z(GB,u,51,[a.a,a.i])}\nfunction d5(a){a.j=wQ;a.k="Paste the text to import into the text area below.";a.b="Accept";sZ(a.I.b,"Paste")}function e5(a){CY();$4.call(this);this.c=a}w(790,784,gQ,e5);_.ci=function(){b5(this)};_.di=function(){ry(this.f,"150px")};_.ei=function(){d5(this)};_.Me=function(){$Y(this);Tq((Qq(),Rq),new f5(this))};_.a=null;_.b=null;_.c=null;function g5(a){CY();e5.call(this,a)}w(789,790,gQ,g5);_.ci=function(){var a;b5(this);a=new S4;Fy(a,new h5(this),(PU(),PU(),QU));this.d=z(GB,u,51,[this.a,a,this.i])};\n_.di=function(){ry(this.f,"150px");EF(new i5(this),this.f)};_.ei=function(){d5(this);this.k+=" Or drag and drop a file on it."};function h5(a){this.a=a}w(791,1,{},h5);_.Md=function(a){var b,c;b=new FileReader;a=(c=nr(a.a),c.files[0]);j5(b,new k5(this));b.readAsText(a)};_.a=null;function k5(a){this.a=a}w(792,1,{},k5);_.ag=function(a){LE();rB(this.a.a.f,a)};_.a=null;function i5(a){this.a=a;this.b=new l5(this);this.c=this.d=1}w(793,568,{},i5);_.a=null;function l5(a){this.a=a}w(794,1,{},l5);\n_.ag=function(a){this.a.a.f.pb[Ik]=null!=a?a:m};_.a=null;function c5(a){this.a=a}w(798,1,{},c5);_.Nd=function(){if(this.a.c){var a=this.a.c,b;b=new GE(a.a,0,dr(this.a.f.pb,Ik));MF(a.a.a,b.a)}JY(this.a,!1)};_.a=null;function f5(a){this.a=a}w(799,1,{},f5);_.pd=function(){Ay(this.a.f.pb,!0);Qz.Qe(this.a.f.pb)};_.a=null;w(800,1,am);_.Ed=function(){var a,b;a=new m5(this.a);void 0!=$wnd.FileReader?b=new g5(a):b=new e5(a);vY(b);X4(b)};function m5(a){this.a=a}w(801,1,{},m5);_.a=null;w(802,1,am);\n_.Ed=function(){var a;a=new $4;var b=this.a,c,d;rB(a.f,b);c=(d=UL(b,"\\r\\n|\\r|\\n|\\n\\r"),d.length);1>=c&&(c=~~(b.length/16));ry(a.f,20*(10>c+1?c+1:10)+uj);Tq((Qq(),Rq),new a5(a));vY(a);X4(a)};function j5(a,b){a.onload=function(a){b.ag(a.target.result)}}X(784);X(790);X(789);X(801);X(787);X(788);X(798);X(799);X(791);X(792);X(793);X(794);X(408);X(457);X(448);X(404);C(fQ)(8);\n//@ sourceURL=8.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/D9A64F1634E29088B910B3E0D4621E49/3.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback3('w(683,677,Ml);_.Ad=function(){this.a.j&&sW(this.a.j);this.a.j=new xW(0,this.a)};B(tO)(3);\n//@ sourceURL=3.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/D9A64F1634E29088B910B3E0D4621E49/4.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback4('w(685,677,Ml);_.Ad=function(){this.a.pc&&sW(this.a.pc);this.a.pc=new xW(1,this.a)};B(tO)(4);\n//@ sourceURL=4.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/D9A64F1634E29088B910B3E0D4621E49/5.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback5('w(686,677,Ml);_.Ad=function(){this.a.y&&(sW(this.a.y),this.a.y=null);0==this.a.ob.v&&(this.a.y=new xW(2,this.a))};B(tO)(5);\n//@ sourceURL=5.js\n') 2 | -------------------------------------------------------------------------------- /static/jsme/jsme/deferredjs/D9A64F1634E29088B910B3E0D4621E49/7.cache.js: -------------------------------------------------------------------------------- 1 | $wnd.jsme.runAsyncCallback7('function d3(a){a=a.Jc.b.c++;return kh+a}w(473,474,yO);_.Je=function(){return d3(this)};_.Ue=function(){this.Tc=new e3((KA(),n),this)};_.$e=function(){return(null==this.Qc&&(this.Qc=d3(this)),this.Qc)+lb+this.Tc.Se()+lb+this.Tc.Te()+lb+this.Yc+Ek+this.Nc+(this.Xc?n:",hidden")+",title="+this.q+(this.p?",resizable":n)+(this.Tc.c.C?",modal":n)};\nfunction f3(){var a,b,c;b=new hV(SO);a=new hV(KO);this.a=new IC(wc,4);new pV;this.c=new g3(this,b);this.c.Tc.c.db=!0;YA(this.c,new dB);QU(this.c,new xV("New atom map"),null);QU(this.c,b,null);QU(this.c,a,null);YA(this.c,new dV(2));c=new cB;QU(c,new yV("atom map",1),null);QU(c,this.a,null);QU(this.c,c,mO);c=new cB;QU(c,b,null);QU(c,a,null);QU(this.c,c,lO);sB(this.c.r,new h3(this))}w(661,1,{},f3);_.a=null;_.b=null;_.c=null;\nfunction g3(a,b){this.a=a;this.b=b;oV.call(this);this.q=n;this.Tc&&this.Tc.Hf(n);this.Tc.c.db=!1;nV(this,!0)}w(662,473,yO,g3);_.Ie=function(a){var b;lV(this);if(Uu(a.f)===Uu(this.b))try{b=wD(Uq(this.a.a.Tc.a.pb,yk),10);var c=this.a.b,d=c.c,e=c.a;tI(d,c.e).b[e].f=b;XE(d,"changeAtomMap",d.o,e,0);GH(d)}catch(f){if(f=vq(f),H(f,102))KV(this.a.b.c,"invalid atom map");else throw f;}return!0};_.a=null;_.b=null;function h3(a){this.a=a}w(663,519,BO,h3);_.zf=function(){lV(this.a.c)};_.a=null;w(676,677,Ml);\n_.Ad=function(){if(this.b.a==(qE(),ue)){!this.a.b&&(this.a.b=new f3);var a=this.a.b,b=this.a,c;if(0Checkpoints 6 | {% endblock %} 7 | 8 | {% block content %} 9 | {{ add_checkpoint('checkpoints', checkpoint_upload_warnings, checkpoint_upload_errors) }} 10 | 11 |
12 | 13 |

Model Checkpoints

14 | 15 | {% for checkpoint in checkpoints %} 16 | 17 | 18 | 22 | 23 | {% endfor %} 24 |
{{checkpoint}} 19 | 20 | 21 |
25 | {% endblock %} 26 | -------------------------------------------------------------------------------- /templates/data.html: -------------------------------------------------------------------------------- 1 | {% extends "layout.html" %} 2 | {% from 'macros.html' import add_data %} 3 | 4 | {% block title %} 5 |

Data

6 | {% endblock %} 7 | 8 | {% block content %} 9 | {{ add_data('data', data_upload_warnings, data_upload_errors) }} 10 | 11 |
12 | 13 |

Datasets

14 | 15 | {% for dataset in datasets %} 16 | 17 | 18 | 22 | 23 | {% endfor %} 24 |
{{dataset}} 19 | 20 | 21 |
25 | 26 | {% endblock %} 27 | -------------------------------------------------------------------------------- /templates/home.html: -------------------------------------------------------------------------------- 1 | {% extends "layout.html" %} 2 | 3 | {% block title %} 4 |

Chemprop — Machine Learning for Molecular Property Prediction

5 | 6 | Message passing 7 | 8 |

Introduction

9 |

This website can be used to predict molecular properties using a Message Passing Neural Network (MPNN). In order to make predictions, an MPNN first needs to be trained on a dataset containing molecules along with known property values for each molecule. Once the MPNN is trained, it can be used to predict those same properties on any new molecules.

10 | 11 |

Training

12 |

To train an MPNN, go to the Train page, upload a dataset or select a dataset which has already been uploaded, set the desired parameters, name the model, and then click "Train".

13 | 14 |

Predicting

15 |

To make property predictions using, go to the Predict page, select the trained model checkpoint you want to use, upload or paste the molecules you would like to make predictions on, and then click "Predict".

16 | 17 |

Managing Data and Checkpoints

18 |

To upload, view, download, or delete datasets and model checkpoints, go to the Data and Checkpoints pages, respectively.

19 | 20 |

Advanced

21 | 22 |

Using GPUs

23 |

If GPUs are available on your machine, you will see a dropdown menu on the Train and Predict pages which will allow you to select a GPU to use. If you select "None" then only CPUs will be used.

24 | 25 |

Working Remotely

26 |

If you wish to train or predict on a remote server, you can use SSH port-forwarding to run training/predicting on the remote server while viewing the website it locally. To do so, follow these instructions:

27 |
    28 |
  1. Connect to the remote server: ssh <remote_user>@<remote_host>
    2. 29 |
  2. On the remote server:
    3. 30 |
      31 |
    1. Navigate to the chemprop directory.
      2. 32 |
    2. Activate the conda environment with the chemprop requirements: source activate <environment_name>
      3. 33 |
    3. Start the website: python web.py
      4. 34 |
    35 |
  3. On your local machine:
    4. 36 |
      37 |
    1. Set up port-forwarding: ssh -N -L 5000:localhost:5000 <remote_user>@<remote_host>
      2. 38 |
    2. Open a web browser and go to localhost:5000
      3. 39 |
    40 |
41 | {% endblock %} 42 | -------------------------------------------------------------------------------- /templates/layout.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 9 | 11 | 14 | 15 | 20 | Chemprop 21 | 22 | 23 | 37 | 38 |
39 | {% block title %}{% endblock %} 40 |
41 | 42 |
43 | {% block content %}{% endblock %} 44 |
45 | 46 |
47 |
48 |

Chemprop v0.1 © 2019

49 | 50 | Source code 51 |
52 |
53 | 54 | 55 | 56 | -------------------------------------------------------------------------------- /templates/macros.html: -------------------------------------------------------------------------------- 1 | {% macro warning_message(message) %} 2 | {% if message %} 3 | 9 | {% endif %} 10 | {% endmacro %} 11 | 12 | {% macro error_message(message) %} 13 | {% if message %} 14 | 20 | {% endif %} 21 | {% endmacro %} 22 | 23 | 24 | {% macro add_data(return_page, data_upload_warnings, data_upload_errors) %} 25 |

Add New Dataset

26 |
27 | 28 |
29 | 30 |
31 | 32 |
33 | 34 | {% if data_upload_warnings %} 35 | {% for warning in data_upload_warnings %} 36 | {{ warning_message(warning) }} 37 | {% endfor %} 38 | {% endif %} 39 | 40 | {% if data_upload_errors %} 41 | {% for error in data_upload_errors %} 42 | {{ error_message(error) }} 43 | {% endfor %} 44 | {% endif %} 45 | {% endmacro %} 46 | 47 | {% macro add_checkpoint(return_page, checkpoint_upload_warnings, checkpoint_upload_errors) %} 48 |

Add New Checkpoint

49 | 50 |
51 | 52 |
53 | 54 |
55 | 56 |
57 | 58 | {% if checkpoint_upload_warnings %} 59 | {% for warning in checkpoint_upload_warnings %} 60 | {{ warning_message(warning) }} 61 | {% endfor %} 62 | {% endif %} 63 | 64 | {% if checkpoint_upload_errors %} 65 | {% for error in checkpoint_upload_errors %} 66 | {{ error_message(error) }} 67 | {% endfor %} 68 | {% endif %} 69 | {% endmacro %} 70 | 71 | {% macro chemdraw() %} 72 | 73 | 78 |
79 | {% endmacro %} 80 | -------------------------------------------------------------------------------- /templates/train.html: -------------------------------------------------------------------------------- 1 | {% extends "layout.html" %} 2 | {% from 'macros.html' import add_data, error_message, warning_message %} 3 | 4 | {% block title %} 5 |

Train

6 | {% endblock %} 7 | 8 | {% block content %} 9 | {{ add_data('train', data_upload_warnings, data_upload_errors) }} 10 | 11 |
12 | 13 |

Train

14 | 15 |
16 |
Data
17 | 22 |
23 |
Dataset type
24 |
25 | 28 | 31 |
32 |
33 | 34 |
Epochs
35 | 36 |
37 |
38 | 39 |
Checkpoint name
40 | 41 | 42 | 43 | {% if cuda %} 44 |
GPU
45 | 51 |
52 |
53 | {% endif %} 54 | 55 |
56 |
57 | 58 | 59 | 60 |
61 | 62 |
63 | 64 | {% if warnings %} 65 | {% for warning in warnings %} 66 | {{ warning_message(warning) }} 67 | {% endfor %} 68 | {% endif %} 69 | 70 | {% if errors %} 71 | {% for error in errors %} 72 | {{ error_message(error) }} 73 | {% endfor %} 74 | {% endif %} 75 | 76 | 90 | 91 | 103 | 104 | 107 | 108 | {% if trained %} 109 |

Training complete!

110 | 111 |

Test performance

112 | 113 |

Overall: {{ mean_score }} {{ metric }}

114 | 115 |
By task
116 | {% for i in range(num_tasks) %} 117 |

{{ task_names[i] }}: {{ task_scores[i] }} {{ metric }}

118 | {% endfor %} 119 | 120 | {% endif %} 121 | 122 | {% endblock %} 123 | -------------------------------------------------------------------------------- /train.py: -------------------------------------------------------------------------------- 1 | import logging 2 | 3 | from chemprop.parsing import parse_train_args 4 | from chemprop.train import cross_validate 5 | from chemprop.utils import set_logger 6 | 7 | 8 | # Initialize logger 9 | logger = logging.getLogger('train') 10 | logger.setLevel(logging.DEBUG) 11 | logger.propagate = False 12 | 13 | if __name__ == '__main__': 14 | args = parse_train_args() 15 | set_logger(logger, args.save_dir, args.quiet) 16 | cross_validate(args, logger) 17 | --------------------------------------------------------------------------------