Skip to content
Get packs

DeepChem Molecular Property Prediction

Guides an agent through safe molecular property ML with DeepChem: featurization, scaffold splits, and validated training.

Source Download skill
Repository
K-Dense-AI/scientific-agent-skills
Stars
55,780
Licence
MIT
Files
10 files

What it does

Once installed, the agent follows a structured DeepChem workflow for predicting chemical properties such as solubility, toxicity or ADMET-style endpoints from SMILES molecular strings. It chooses matching featurizers for each model type, uses scaffold or grouped holdouts so test molecules differ from training ones, keeps missing labels masked, and reports metrics in original units. It also warns when results are not evidence of real-world performance.

Great for

  • Chemistry or cheminformatics researchers building baseline property models
  • Data scientists learning MoleculeNet benchmarks and molecular featurization
  • Teams that need careful validation rather than a quick accuracy number

Skip it if

  • Regulatory, clinical or safety decisions based only on a trained model, since the skill itself states that successful fitting is not prospective evidence
  • Users without Python 3.11 and RDKit who need a turnkey app, because setup is nontrivial and GPU or graph-model backends (DGL, DGL-LifeSci) were not verified
  • Non-chemistry tabular ML, where the skill's molecular contracts do not apply

Try asking

  • “Train a solubility model on my CSV of SMILES and measured logS values using a scaffold split”
  • “Run the ESOL benchmark and compare a fingerprint baseline against a Torch multitask regressor”
  • “Fine-tune a ChemBERTa-style model on the BBBP dataset for one binary task”
  • “Predict solubility for the SMILES CCO and benzene after training on my data”
Step-by-step guide: how to use DeepChem Molecular Property PredictionInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 10 original files, the MIT licence and an attribution note.

    deepchem.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip deepchem.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip deepchem.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Use the bundled scripts on a copy of your data and save split IDs, rejected rows and predictions yourself, since the scripts only print results.
  • Install in an isolated Python 3.11 environment and check that the model path you need (e.g. DGL for GCN) actually imports before relying on it.
  • Ask the agent to report per-task support, baseline comparisons and applicability limits alongside any metric.

What's inside

  • references/api_reference.md
  • references/core_capabilities.md
  • references/review.md
  • references/typical_workflows.md
  • references/workflows.md
  • scripts/_common.py
  • scripts/graph_neural_network.py
  • scripts/predict_solubility.py
  • scripts/transfer_learning.py
  • SKILL.md

Redistributed unmodified from K-Dense-AI/scientific-agent-skills under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.