Deepchem

A batteries-included molecular ML library that shines when you need fast property prediction experiments without building everything from scratch. Ships with 30+ MoleculeNet benchmark datasets (Tox21, BBBP, Delaney) and extensive featurizers covering fingerprints, graph representations, and descriptors. The GNN implementations (GCN, GAT, AttentiveFP) work well for larger datasets, while traditional ML wrappers handle smaller ones. Real strength is the ecosystem: loaders for SMILES, SDF, and FASTA files, scaffold-based splitting to prevent data leakage, and pretrained models for transfer learning. Note the Python 3.11 cap and backend installation quirks (install PyTorch or TensorFlow first). If you want a more PyTorch-native graph workflow, consider torchdrug instead.

skills.sh installs: 2K
GitHub stars: 56.6K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add k-dense-ai/scientific-agent-skills --skill deepchem --agent claude-code

Installs into .claude/skills of the current project.

Files

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.