deepchem

Name: deepchem
Author: K-Dense-AI

31kstars3.1kforks

How to install deepchem?

k-dense-ai/scientific-agent-skills/deepchem

$npx -y skills add K-Dense-AI/scientific-agent-skills --skill deepchem

Installs into the current project.

Summary

Molecular ML with diverse featurizers and pre-built datasets. Use for property prediction (ADMET, toxicity) with traditional ML or GNNs when you want extensive featurization options and MoleculeNet benchmarks. Best for quick experiments with pre-trained models, diverse molecular

Files · 1

View on GitHub

SKILL.md

1	# DeepChem
2
3	## Overview
4
5	DeepChem is a comprehensive Python library for applying machine learning to chemistry, materials science, and biology. Enable molecular property prediction, drug discovery, materials design, and biomolecule analysis through specialized neural networks, molecular featurization methods, and pretrained models.
6
7	Version note: Examples target deepchem 2.8.0 (PyPI stable, Apr 2024). Requires Python 3.7–3.11 (`<3.12` on PyPI). Core utilities (loaders, featurizers, MoleculeNet) work without a DL backend; GNN and transformer models need the matching extra (`torch`, `tensorflow`, or `jax`). Install the backend framework first when using GPU builds.
8
9	## When to Use This Skill
10
11	This skill should be used when:
12	- Loading and processing molecular data (SMILES strings, SDF files, protein sequences)
13	- Predicting molecular properties (solubility, toxicity, binding affinity, ADMET properties)
14	- Training models on chemical/biological datasets
15	- Using MoleculeNet benchmark datasets (Tox21, BBBP, Delaney, etc.)
16	- Converting molecules to ML-ready features (fingerprints, graph representations, descriptors)
17	- Implementing graph neural networks for molecules (GCN, GAT, MPNN, AttentiveFP)
18	- Applying transfer learning with pretrained models (ChemBERTa, GROVER, MolFormer)
19	- Predicting crystal/materials properties (bandgap, formation energy)
20	- Analyzing protein or DNA sequences
21
22	## Core Capabilities
23
24	### 1. Molecular Data Loading and Processing
25
26	DeepChem provides specialized loaders for various chemical data formats:
27
28	```python
29	import deepchem as dc
30
31	# Load CSV with SMILES
32	featurizer = dc.feat.CircularFingerprint(radius=2, size=2048)
33	loader = dc.data.CSVLoader(
34	tasks=['solubility', 'toxicity'],
35	feature_field='smiles',
36	featurizer=featurizer
37	)
38	dataset = loader.create_dataset('molecules.csv')
39
40	# Load SDF files
41	loader = dc.data.SDFLoader(tasks=['activity'], featurizer=featurizer)
42	dataset = loader.create_dataset('compounds.sdf')
43
44	# Load protein sequences
45	loader = dc.data.FASTALoader()
46	dataset = loader.create_dataset('proteins.fasta')
47	```
48
49	Key Loaders:
50	- `CSVLoader`: Tabular data with molecular identifiers
51	- `SDFLoader`: Molecular structure files
52	- `FASTALoader`: Protein/DNA sequences
53	- `ImageLoader`: Molecular images
54	- `JsonLoader`: JSON-formatted datasets
55
56	### 2. Molecular Featurization
57
58	Convert molecules into numerical representations for ML models.
59
60	#### Decision Tree for Featurizer Selection
61
62	```
63	Is the model a graph neural network?
64	├─ YES → Use graph featurizers
65	│ ├─ Standard GNN → MolGraphConvFeaturizer
66	│ ├─ Message passing → DMPNNFeaturizer
67	│ └─ Pretrained → GroverFeaturizer
68	│
69	└─ NO → What type of model?
70	├─ Traditional ML (RF, XGBoost, SVM)
71	│ ├─ Fast baseline → CircularFingerprint (ECFP)
72	│ ├─ Interpretable → RDKitDescriptors
73	│ └─ Maximum coverage → MordredDescriptors
74	│
75	├─ Deep learning (non-graph)
76	│ ├─ Dense networks → CircularFingerprint
77	│ └─ CNN → SmilesToImage
78	│
79	├─ Sequence models (LSTM, Transformer)
80	│ └─ SmilesToSeq
81	│
82	└─ 3D structure analysis
83	└─ CoulombMatrix
84	```
85
86	#### Example Featurization
87
88	```python
89	# Fingerprints (for traditional ML)
90	fp = dc.feat.CircularFingerprint(radius=2, size=2048)
91
92	# Descriptors (for interpretable models)
93	desc = dc.feat.RDKitDescriptors()
94
95	# Graph features (for GNNs)
96	graph_feat = dc.feat.MolGraphConvFeaturizer()
97
98	# Apply featurization
99	features = fp.featurize(['CCO', 'c1ccccc1'])
100	```
101
102	Selection Guide:
103	- Small datasets (<1K): CircularFingerprint or RDKitDescriptors
104	- Medium datasets (1K-100K): CircularFingerprint or graph featurizers
105	- Large datasets (>100K): Graph featurizers (MolGraphConvFeaturizer, DMPNNFeaturizer)
106	- Transfer learning: Pretrained model featurizers (GroverFeaturizer)
107
108	See `references/api_reference.md` for complete featurizer documentation.
109
110	### 3. Data Splitting
111
112	Critical: For drug discovery tasks, use `ScaffoldSplitter` to prevent data leakage from similar molecular structures appearing in both training and test sets.
113
114	```python
115	# Scaffold splitting (recommended for molecules)
116	splitter = dc.splits.ScaffoldSplitter()
117	train, valid, test = splitter.train_valid_test_split(
118	dataset,
119	frac_train=0.8,
120	frac_valid=0.1,
121	frac_test=0.1
122	)
123
124	# Ra

People who liked this also viewed

6 picks

Type

deepchem

31kstars3.1kforks

View on GitHub

How to install deepchem?

k-dense-ai/scientific-agent-skills/deepchem

$npx -y skills add K-Dense-AI/scientific-agent-skills --skill deepchem

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# DeepChem
2
3	## Overview
4
5	DeepChem is a comprehensive Python library for applying machine learning to chemistry, materials science, and biology. Enable molecular property prediction, drug discovery, materials design, and biomolecule analysis through specialized neural networks, molecular featurization methods, and pretrained models.
6
7	Version note: Examples target deepchem 2.8.0 (PyPI stable, Apr 2024). Requires Python 3.7–3.11 (`<3.12` on PyPI). Core utilities (loaders, featurizers, MoleculeNet) work without a DL backend; GNN and transformer models need the matching extra (`torch`, `tensorflow`, or `jax`). Install the backend framework first when using GPU builds.
8
9	## When to Use This Skill
10
11	This skill should be used when:
12	- Loading and processing molecular data (SMILES strings, SDF files, protein sequences)
13	- Predicting molecular properties (solubility, toxicity, binding affinity, ADMET properties)
14	- Training models on chemical/biological datasets
15	- Using MoleculeNet benchmark datasets (Tox21, BBBP, Delaney, etc.)
16	- Converting molecules to ML-ready features (fingerprints, graph representations, descriptors)
17	- Implementing graph neural networks for molecules (GCN, GAT, MPNN, AttentiveFP)
18	- Applying transfer learning with pretrained models (ChemBERTa, GROVER, MolFormer)
19	- Predicting crystal/materials properties (bandgap, formation energy)
20	- Analyzing protein or DNA sequences
21
22	## Core Capabilities
23
24	### 1. Molecular Data Loading and Processing
25
26	DeepChem provides specialized loaders for various chemical data formats:
27
28	```python
29	import deepchem as dc
30
31	# Load CSV with SMILES
32	featurizer = dc.feat.CircularFingerprint(radius=2, size=2048)
33	loader = dc.data.CSVLoader(
34	tasks=['solubility', 'toxicity'],
35	feature_field='smiles',
36	featurizer=featurizer
37	)
38	dataset = loader.create_dataset('molecules.csv')
39
40	# Load SDF files
41	loader = dc.data.SDFLoader(tasks=['activity'], featurizer=featurizer)
42	dataset = loader.create_dataset('compounds.sdf')
43
44	# Load protein sequences
45	loader = dc.data.FASTALoader()
46	dataset = loader.create_dataset('proteins.fasta')
47	```
48
49	Key Loaders:
50	- `CSVLoader`: Tabular data with molecular identifiers
51	- `SDFLoader`: Molecular structure files
52	- `FASTALoader`: Protein/DNA sequences
53	- `ImageLoader`: Molecular images
54	- `JsonLoader`: JSON-formatted datasets
55
56	### 2. Molecular Featurization
57
58	Convert molecules into numerical representations for ML models.
59
60	#### Decision Tree for Featurizer Selection
61
62	```
63	Is the model a graph neural network?
64	├─ YES → Use graph featurizers
65	│ ├─ Standard GNN → MolGraphConvFeaturizer
66	│ ├─ Message passing → DMPNNFeaturizer
67	│ └─ Pretrained → GroverFeaturizer
68	│
69	└─ NO → What type of model?
70	├─ Traditional ML (RF, XGBoost, SVM)
71	│ ├─ Fast baseline → CircularFingerprint (ECFP)
72	│ ├─ Interpretable → RDKitDescriptors
73	│ └─ Maximum coverage → MordredDescriptors
74	│
75	├─ Deep learning (non-graph)
76	│ ├─ Dense networks → CircularFingerprint
77	│ └─ CNN → SmilesToImage
78	│
79	├─ Sequence models (LSTM, Transformer)
80	│ └─ SmilesToSeq
81	│
82	└─ 3D structure analysis
83	└─ CoulombMatrix
84	```
85
86	#### Example Featurization
87
88	```python
89	# Fingerprints (for traditional ML)
90	fp = dc.feat.CircularFingerprint(radius=2, size=2048)
91
92	# Descriptors (for interpretable models)
93	desc = dc.feat.RDKitDescriptors()
94
95	# Graph features (for GNNs)
96	graph_feat = dc.feat.MolGraphConvFeaturizer()
97
98	# Apply featurization
99	features = fp.featurize(['CCO', 'c1ccccc1'])
100	```
101
102	Selection Guide:
103	- Small datasets (<1K): CircularFingerprint or RDKitDescriptors
104	- Medium datasets (1K-100K): CircularFingerprint or graph featurizers
105	- Large datasets (>100K): Graph featurizers (MolGraphConvFeaturizer, DMPNNFeaturizer)
106	- Transfer learning: Pretrained model featurizers (GroverFeaturizer)
107
108	See `references/api_reference.md` for complete featurizer documentation.
109
110	### 3. Data Splitting
111
112	Critical: For drug discovery tasks, use `ScaffoldSplitter` to prevent data leakage from similar molecular structures appearing in both training and test sets.
113
114	```python
115	# Scaffold splitting (recommended for molecules)
116	splitter = dc.splits.ScaffoldSplitter()
117	train, valid, test = splitter.train_valid_test_split(
118	dataset,
119	frac_train=0.8,
120	frac_valid=0.1,
121	frac_test=0.1
122	)
123
124	# Ra

People who liked this also viewed

6 picks

Type