DeepChem

When to use

Use for molecular property prediction, SMILES/graph featurization, MoleculeNet benchmarks and explicitly configured encoder transfer. The workflow also covers DeepChem materials and sequence adapters when their data/model contracts are met.

Targets DeepChem 2.8.0, still the stable PyPI release at review. Nightly 2.8.1 builds have additional APIs and different constraints; do not mix latest docs with a stable installation. The tested CPU stack and optional-backend limitations are in references/review.md.

Workflow

1. Define compound identity, assay conditions, target units and missing labels. Parse SMILES and retain a record of rejected rows. Audit duplicate structures and repeated measurements before splitting.
2. Match the representation to the model using the table below. Fit a numeric baseline first; dataset size alone does not select the best architecture.
3. Choose a scaffold, temporal or grouped holdout for the deployment question. Preserve exact split IDs, inspect scaffold overlap and per-task class support, and reject empty splits. Scaffold splitting does not eliminate all leakage.
4. Fit preprocessing on training data only. Preserve dataset.w masks. Normalize continuous targets only, and invert those transforms for metrics/predictions.
5. Select model settings/epochs on validation data and evaluate the final holdout once. Report per-task support, uncertainty across planned repeats and baseline comparisons. AUC requires both observed classes.
6. Predict using the training representation and transforms. Preserve identifiers, original target units and applicability limits. Successful fitting is not evidence of prospective scientific performance.

Essential model contracts

MoleculeNet's 'ECFP' alias is 1024 bits, while the fingerprint class defaults to 2048. 'GraphConv' is legacy ConvMolFeaturizer, incompatible with Torch GCN. 'Raw' defaults to RDKit Mol objects; use DummyFeaturizer() for raw strings. Import Torch MPNN/GROVER from deepchem.models.torch_models.

Stable HuggingFaceModel returns logits and its training loss ignores sample weights. The bundled HF transfer script accepts one fully observed, unweighted binary/regression task and rejects sparse multitask datasets. A model ID or model_dir is not itself a loaded pretrained model.

Installation

Use an isolated Python 3.11 environment, outside any repository environment whose Python requirement conflicts. This tested core stack supports the fingerprint, DMPNN, local HF and GROVER smoke paths:

uv venv --python 3.11 .venv-deepchem
uv pip install --python .venv-deepchem/bin/python \
  'deepchem==2.8.0' 'torch==2.14.1' 'transformers==5.18.0' \
  'torch-geometric==2.8.0.post1'

The CPU smoke checks used these versions, not GPU builds. For GPU training install the correct framework build according to its official platform instructions. GCN/GAT/AttentiveFP/Torch MPNN additionally require mutually compatible DGL and DGL-LifeSci; those backends were not available in this audit. Installing Torch alone, or importing DeepChem successfully, does not establish those model paths.

The stable distribution exposes torch, tensorflow, jax, and dqc extras, with historical optional dependency requirements; they are not an assurance that every modern platform/backend combination resolves or executes. No [all] extra exists. The TensorFlow, JAX and quantum-chemistry routes were source-reviewed only. Check release installation docs and requirements. DeepChem attempts optional imports at package import time and can log skipped modules; it does not promise all classes are available after that import.

Bundled scripts

The primary executable end-to-end example is predict_solubility.py with a custom

…