Alterlab Deepchem

---
name: alterlab-deepchem
description: Runs molecular machine learning with DeepChem — diverse featurizers, pre-built MoleculeNet benchmark datasets, and pre-trained models (ChemBERTa, GROVER) for property prediction (ADMET, toxicity, solubility) via traditional ML or graph neural networks. Use when running end-to-end molecular ML experiments that need MoleculeNet benchmarks, scaffold splitting, or ready-made models with minimal setup; for building custom PyTorch graph architectures prefer alterlab-torch-geometric, and for standalone molecule-to-feature-vector generation prefer alterlab-molfeat. Part of the AlterLab Academic Skills suite.
license: MIT
allowed-tools: Read Write Edit Bash(python:*) Bash(uv:*)
compatibility: "Self-contained, no API key or account. Use a dedicated environment: DeepChem 2.8.0 needs Python <= 3.11 (nightly --pre builds: <= 3.12, numpy<2, torch==2.2.1 with the [torch] extra)."
metadata:
    skill-author: AlterLab
    version: "1.1.0"
    last_updated: "2026-09-23"
---

# DeepChem

## Overview

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.

## When to Use This Skill

This skill should be used when:
- Loading and processing molecular data (SMILES strings, SDF files, protein sequences)
- Predicting molecular properties (solubility, toxicity, binding affinity, ADMET properties)
- Training models on chemical/biological datasets
- Using MoleculeNet benchmark datasets (Tox21, BBBP, Delaney, etc.)
- Converting molecules to ML-ready features (fingerprints, graph representations, descriptors)
- Implementing graph neural networks for molecules (GCN, GAT, MPNN, AttentiveFP)
- Applying transfer learning with pretrained models (ChemBERTa, GROVER; MoLFormer in nightly builds)
- Predicting crystal/materials properties (bandgap, formation energy)
- Analyzing protein or DNA sequences

### Does NOT Trigger

| Scenario | Use Instead |
|----------|-------------|
| Designing a custom GNN architecture (own layers / message passing) in PyTorch | `alterlab-torch-geometric` (or `alterlab-torchdrug` for TorchDrug's task abstractions) |
| Only fetching standardized ADMET/DTI benchmark datasets and official splits, no modeling | `alterlab-pytdc` |
| Only turning molecules into feature matrices or pretrained embeddings for your own models | `alterlab-molfeat` |
| Classical ML on an existing numeric feature table (pipelines, CV, tuning) | `alterlab-scikit-learn` |

## Core Capabilities

### 1. Molecular Data Loading and Processing

DeepChem provides specialized loaders for various chemical data formats:

```python
import deepchem as dc

# Load CSV with SMILES
featurizer = dc.feat.CircularFingerprint(radius=2, size=2048)
loader = dc.data.CSVLoader(
    tasks=['solubility', 'toxicity'],
    feature_field='smiles',
    featurizer=featurizer
)
dataset = loader.create_dataset('molecules.csv')

# Load SDF files
loader = dc.data.SDFLoader(tasks=['activity'], featurizer=featurizer)
dataset = loader.create_dataset('compounds.sdf')

# Load protein sequences
loader = dc.data.FASTALoader()
dataset = loader.create_dataset('proteins.fasta')
```

**Key Loaders**:
- `CSVLoader`: Tabular data with molecular identifiers
- `SDFLoader`: Molecular structure files
- `FASTALoader`: Protein/DNA sequences
- `ImageLoader`: Molecular images
- `JsonLoader`: JSON-formatted datasets

### 2. Molecular Featurization

Convert molecules into numerical representations for ML models.

#### Decision Tree for Featurizer Selection

```
Is the model a graph neural network?
├─ YES → Use graph featurizers
│   ├─ Standard GNN → MolGraphConvFeaturizer
│   ├─ Message passing → DMPNNFeaturizer
│   └─ Pretrained → GroverFeaturizer
│
└─ NO → What type of model?
    ├─ Traditional ML (RF, XGBoost, SVM)
    │   ├─ Fast baseline → CircularFingerprint (ECFP)
    │   ├─ Interpretable → RDKitDescriptors
    │   └─ Maximum coverage → MordredDescriptors
    │
    ├─ Deep learning (non-graph)
    │   ├─ Dense networks → CircularFingerprint
    │   └─ CNN → SmilesToImage
    │
    ├─ Sequence models (LSTM, Transformer)
    │   └─ SmilesToSeq
    │
    └─ 3D structure analysis
        └─ CoulombMatrix
```

#### Example Featurization

```python
# Fingerprints (for traditional ML)
fp = dc.feat.CircularFingerprint(radius=2, size=2048)

# Descriptors (for interpretable models)
desc = dc.feat.RDKitDescriptors()

# Graph features (for GNNs)
graph_feat = dc.feat.MolGraphConvFeaturizer()

# Apply featurization
features = fp.featurize(['CCO', 'c1ccccc1'])
```

**Selection Guide**:
- **Small datasets (<1K)**: CircularFingerprint or RDKitDescriptors
- **Medium datasets (1K-100K)**: CircularFingerprint or graph featurizers
- **Large datasets (>100K)**: Graph featurizers (MolGraphConvFeaturizer, DMPNNFeaturizer)
- **Transfer learning**: Pretrained model featurizers (GroverFeaturizer)

See `references/api_reference.md` for complete featurizer documentation.

### 3. Data Splitting

**Critical**: For drug discovery tasks, use `ScaffoldSplitter` to prevent data leakage from similar molecular structures appearing in both training and test sets.

```python
# Scaffold splitting (recommended for molecules)
splitter = dc.splits.ScaffoldSplitter()
train, valid, test = splitter.train_valid_test_split(
    dataset,
    frac_train=0.8,
    frac_valid=0.1,
    frac_test=0.1
)

# Random splitting (for non-molecular data)
splitter = dc.splits.RandomSplitter()
train, test = splitter.train_test_split(dataset)

# Stratified splitting (for imbalanced classification)
splitter = dc.splits.RandomStratifiedSplitter()
train, test = splitter.train_test_split(dataset)
```

**Available Splitters**:
- `ScaffoldSplitter`: Split by molecular scaffolds (prevents leakage)
- `ButinaSplitter`: Clustering-based molecular splitting
- `MaxMinSplitter`: Maximize diversity between sets
- `RandomSplitter`: Random splitting
- `RandomStratifiedSplitter`: Preserves class distributions

### 4. Model Selection and Training

#### Quick Model Selection Guide

| Dataset Size | Task | Recommended Model | Featurizer |
|-------------|------|-------------------|------------|
| < 1K samples | Any | SklearnModel (RandomForest) | CircularFingerprint |
| 1K-100K | Classification/Regression | GBDTModel or MultitaskRegressor | CircularFingerprint |
| > 100K | Molecular properties | GCNModel, AttentiveFPModel, DMPNNModel | MolGraphConvFeaturizer |
| Any (small preferred) | Transfer learning | Chemberta, GroverModel (MoLFormer: nightly only) | Model-specific |
| Crystal structures | Materials properties | CGCNNModel, MEGNetModel | Structure-based |
| Protein sequences | Protein properties | ProtBERT (nightly builds only) | Sequence-based |

#### Example: Traditional ML
```python
from sklearn.ensemble import RandomForestRegressor

# Wrap scikit-learn model
sklearn_model = RandomForestRegressor(n_estimators=100)
model = dc.models.SklearnModel(model=sklearn_model)
model.fit(train)
```

#### Example: Deep Learning
```python
# Multitask regressor (for fingerprints)
model = dc.models.MultitaskRegressor(
    n_tasks=2,
    n_features=2048,
    layer_sizes=[1000, 500],
    dropouts=0.25,
    learning_rate=0.001
)
model.fit(train, nb_epoch=50)
```

#### Example: Graph Neural Networks
```python
# Graph Convolutional Network
model = dc.models.GCNModel(
    n_tasks=1,
    mode='regression',
    batch_size=128,
    learning_rate=0.001
)
model.fit(train, nb_epoch=50)

# Graph Attention Network
model = dc.models.GATModel(n_tasks=1, mode='classification')
model.fit(train, nb_epoch=50)

# Attentive Fingerprint
model = dc.models.AttentiveFPModel(n_tasks=1, mode='regression')
model.fit(train, nb_epoch=50)
```

### 5. MoleculeNet Benchmarks

Quick access to 30+ curated benchmark datasets with standardized train/valid/test splits:

```python
# Load benchmark dataset
tasks, datasets, transformers = dc.molnet.load_tox21(
    featurizer='GraphConv',  # or 'ECFP', 'Weave', 'Raw'
    splitter='scaffold',     # or 'random', 'stratified'
    reload=False
)
train, valid, test = datasets

# Train and evaluate
model = dc.models.GCNModel(n_tasks=len(tasks), mode='classification')
model.fit(train, nb_epoch=50)

metric = dc.metrics.Metric(dc.metrics.roc_auc_score)
test_score = model.evaluate(test, [metric])
```

**Common Datasets**:
- **Classification**: `load_tox21()`, `load_bbbp()`, `load_hiv()`, `load_clintox()`
- **Regression**: `load_delaney()`, `load_freesolv()`, `load_lipo()`
- **Quantum properties**: `load_qm7()`, `load_qm8()`, `load_qm9()`
- **Materials**: `load_perovskite()`, `load_bandgap()`, `load_mp_formation_energy()`

See `references/api_reference.md` for complete dataset list.

### 6. Transfer Learning

Leverage pretrained models for improved performance, especially on small datasets:

```python
# ChemBERTa (RoBERTa pretrained on SMILES) — use DeepChem's Chemberta wrapper
model = dc.models.Chemberta(
    task='classification',
    tokenizer_path='seyonec/PubChem10M_SMILES_BPE_60k',
    n_tasks=1,
    learning_rate=2e-5  # passed via **kwargs to TorchModel
)
model.fit(train, nb_epoch=10)

# GROVER: DeepChem's GroverModel is NOT a one-line pretrained loader. task is
# 'pretraining' | 'finetuning', mode is 'classification' | 'regression', and it needs
# node/edge feature dims, atom/bond vocabularies built from your data, and inputs
# featurized with dc.feat.GroverFeaturizer(features_generator=dc.feat.CircularFingerprint()).
# No pretrained GROVER weights are downloaded — pretrain (task='pretraining') or restore
# your own checkpoint. Full recipe: the GroverModel class docstring.
from deepchem.feat.vocabulary_builders import (GroverAtomVocabularyBuilder,
                                               GroverBondVocabularyBuilder)
av, bv = GroverAtomVocabularyBuilder(), GroverBondVocabularyBuilder()
av.build(smiles_dataset); bv.build(smiles_dataset)   # dataset built with DummyFeaturizer
model = dc.models.GroverModel(node_fdim=151, edge_fdim=165, hidden_size=128,
                              atom_vocab=av, bond_vocab=bv, features_dim=2048,
                              functional_group_size=85, task='finetuning',
                              mode='regression', n_tasks=1, model_dir='grover_ft')
model.fit(grover_train, nb_epoch=20)   # grover_train featurized with GroverFeaturizer
```

**When to use transfer learning**:
- Small datasets (< 1000 samples)
- Novel molecular scaffolds
- Limited computational resources
- Need for rapid prototyping

Use the `scripts/transfer_learning.py` script for guided transfer learning workflows.

### 7. Model Evaluation

```python
# Define metrics
classification_metrics = [
    dc.metrics.Metric(dc.metrics.roc_auc_score, name='ROC-AUC'),
    dc.metrics.Metric(dc.metrics.accuracy_score, name='Accuracy'),
    dc.metrics.Metric(dc.metrics.f1_score, name='F1')
]

regression_metrics = [
    dc.metrics.Metric(dc.metrics.r2_score, name='R²'),
    dc.metrics.Metric(dc.metrics.mean_absolute_error, name='MAE'),
    dc.metrics.Metric(dc.metrics.rms_score, name='RMSE')   # there is no root_mean_squared_error
]

# Evaluate
train_scores = model.evaluate(train, classification_metrics)
test_scores = model.evaluate(test, classification_metrics)
```

### 8. Making Predictions

```python
# Predict on test set
predictions = model.predict(test)

# Predict on new molecules
new_smiles = ['CCO', 'c1ccccc1', 'CC(C)O']
new_features = featurizer.featurize(new_smiles)
new_dataset = dc.data.NumpyDataset(X=new_features)

# Pass the training transformers to predict() to undo y-normalization
# so predictions come back in the original units. (Don't .transform()
# the X-only dataset — those transformers act on y, not the features.)
predictions = model.predict(new_dataset, transformers=transformers)
```

## Typical Workflows

Three ready-to-run end-to-end recipes are provided:
- **Workflow A — Quick Benchmark Evaluation**: load a MoleculeNet benchmark, train a GNN, and score it.
- **Workflow B — Custom Data Prediction**: featurize a CSV, scaffold-split, normalize, train, and evaluate.
- **Workflow C — Transfer Learning on a Small Dataset**: fine-tune a pretrained model on raw SMILES.

Full runnable code for A/B/C: see `references/end_to_end_recipes.md`. For eight deeper workflows (molecular generation, materials science, protein analysis, custom-model integration, hyperparameter search): see `references/workflows.md`.

## Example Scripts

This skill includes three production-ready scripts in the `scripts/` directory:

### 1. `predict_solubility.py`
Train and evaluate solubility prediction models. Works with Delaney benchmark or custom CSV data.

```bash
# Use Delaney benchmark
python scripts/predict_solubility.py

# Use custom data
python scripts/predict_solubility.py \
    --data my_data.csv \
    --smiles-col smiles \
    --target-col solubility \
    --predict "CCO" "c1ccccc1"
```

### 2. `graph_neural_network.py`
Train various graph neural network architectures on molecular data.

```bash
# Train GCN on Tox21
python scripts/graph_neural_network.py --model gcn --dataset tox21

# Train AttentiveFP on custom data
python scripts/graph_neural_network.py \
    --model attentivefp \
    --data molecules.csv \
    --task-type regression \
    --targets activity \
    --epochs 100
```

### 3. `transfer_learning.py`
Fine-tune pretrained ChemBERTa on molecular property prediction tasks. (`--model grover` exits with setup guidance: GROVER needs the manual vocabulary/featurizer recipe shown above.)

```bash
# Fine-tune ChemBERTa on BBBP
python scripts/transfer_learning.py --model chemberta --dataset bbbp

# Fine-tune ChemBERTa on custom data
python scripts/transfer_learning.py \
    --model chemberta \
    --data small_dataset.csv \
    --target activity \
    --task-type classification \
    --epochs 20
```

## Common Patterns, Best Practices, and Pitfalls

Core habits: always scaffold-split molecular data to prevent leakage; normalize features and targets; start simple (Random Forest + `CircularFingerprint`) before scaling to deep nets and GNNs; balance imbalanced data with `BalancingTransformer` or balanced metrics; and use `DiskDataset` with smaller batch sizes to avoid memory issues. Recurring failure modes — data leakage, GNNs underperforming fingerprints, overfitting on small datasets, and import errors — each have concrete fixes.

Full pattern recipes (with code) and the pitfall-to-fix catalog: see `references/best_practices.md`.

## Reference Documentation

This skill includes comprehensive reference documentation:

### `references/api_reference.md`
Complete API documentation including:
- All data loaders and their use cases
- Dataset classes and when to use each
- Complete featurizer catalog with selection guide
- Model catalog organized by category (50+ models)
- MoleculeNet dataset descriptions
- Metrics and evaluation functions
- Common code patterns

**When to reference**: Search this file when you need specific API details, parameter names, or want to explore available options.

### `references/workflows.md`
Eight detailed end-to-end workflows:
1. Molecular property prediction from SMILES
2. Using MoleculeNet benchmarks
3. Hyperparameter optimization
4. Transfer learning with pretrained models
5. Molecular generation with GANs
6. Materials property prediction
7. Protein sequence analysis
8. Custom model integration

**When to reference**: Use these workflows as templates for implementing complete solutions.

### `references/end_to_end_recipes.md`
Three quick end-to-end recipes (benchmark evaluation, custom-data prediction, transfer learning) with full runnable code.

**When to reference**: Grab one of these as a starting scaffold for a complete pipeline.

### `references/best_practices.md`
Best-practice patterns (splitting, normalization, model progression, class balancing, memory) and a pitfall-to-fix troubleshooting catalog.

**When to reference**: Consult when debugging poor performance or choosing a modeling strategy.

## Installation Notes

Give DeepChem its own virtual environment — its pins conflict with current scientific stacks:

```bash
# Stable release 2.8.0 (Apr 2024; latest stable as of 2026-09) supports Python 3.7-3.11
uv venv --python 3.11 && uv pip install deepchem

# PyTorch-backed models (GCN, GAT, AttentiveFP, Chemberta, GroverModel, ...)
uv pip install "deepchem[torch]"       # also: "deepchem[tensorflow]", "deepchem[jax]" — there is no [all] extra

# Python 3.12 or newer model classes (MoLFormer, ProtBERT): nightly pre-releases
uv pip install --pre "deepchem[torch]" # 2.8.1.devYYYYMMDD builds; Python <= 3.12
```

The nightlies pin `numpy<2` and, with `[torch]`, `torch==2.2.1` and `dgl<2.2.1`, so they cannot share an environment with numpy 2.x / current PyTorch. The Hugging Face wrappers (`Chemberta`, `MoLFormer`, `ProtBERT`) also need `transformers`, which no extra installs.

## Additional Resources

- Official documentation: https://deepchem.readthedocs.io/
- GitHub repository: https://github.com/deepchem/deepchem
- Tutorials: https://deepchem.readthedocs.io/en/latest/get_started/tutorials.html
- Paper: "MoleculeNet: A Benchmark for Molecular Machine Learning"

Part of the AlterLab Academic Skills suite.