skills.rest
silverstein
silversteinCOMMUNITY
Explore all skills by this author →
Website𝕏@matsilversteinGitHub

deepchem

Build molecular machine learning workflows for property prediction from SMILES data.

21|2|Updated Dec 8, 2025
Search Tags:#deep learning#graph neural networks#scaffold splitting#transfer learning#molecular property prediction#SMILES featurization
One-click install
npx skills add https://github.com/silverstein/claude-scientific-skills-desktop --skill deepchem-silverstein
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: deepchem
Source: https://github.com/silverstein/claude-scientific-skills-desktop/tree/main/corpus/deepchem
Command: npx skills add https://github.com/silverstein/claude-scientific-skills-desktop --skill deepchem-silverstein

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires deepchem, numpy, python3, argparse, sys, and includes scripts (resource) and references (resource) components.

What problem does it solve?

DeepChem skill helps you build reliable machine learning pipelines for molecular property prediction from SMILES or molecular files, including featurization, splitting, training, evaluation, and transfer learning.

Core Features & Use Cases

  • Load and preprocess molecular data from CSV/SMILES, SDF, and biological FASTA inputs for tasks like activity, toxicity, and property prediction.
  • Choose the right featurization strategy for classic ML (fingerprints/descriptors), graph neural networks (graph featurizers and GNN models), and transformer-style transfer learning (raw SMILES with pretrained models).
  • Train and evaluate models safely using scaffold-based splitting to reduce data leakage, then compare models with appropriate metrics (ROC-AUC/accuracy/F1 for classification; R²/MAE/RMSE for regression).

Quick Start

Ask the AI to help you train a solubility predictor by running the script on the Delaney (ESOL) benchmark and reporting test metrics.

Frequently Asked Questions about deepchem

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I predict molecular properties from SMILES strings using deep learning?▼

To predict molecular properties from SMILES, you can use DeepChem loaders to preprocess data, apply featurization strategies like fingerprints or graph featurizers, and train models for regression or classification tasks.

What is scaffold splitting and why is it needed for molecular machine learning?▼

Scaffold splitting is a data partitioning method that separates molecules by their structural backbones. It is needed to reduce data leakage and ensure reliable evaluation when training molecular machine learning models.

Can I use transfer learning with pretrained models like ChemBERTa for molecular property prediction?▼

Yes, you can use transfer learning for molecular property prediction by taking raw SMILES data and applying pretrained transformer-style models like ChemBERTa or GROVER to leverage learned chemical representations.

How do I choose the right featurization strategy for graph neural networks?▼

Choosing the right featurization strategy depends on your model: use fingerprints and descriptors for classic ML, and use graph featurizers specifically designed for graph neural networks to represent molecular structures.

What evaluation metrics should I use for classification and regression in MoleculeNet benchmarks?▼

For MoleculeNet benchmarks, use ROC-AUC, accuracy, and F1 metrics for classification tasks, while using R², MAE, and RMSE metrics to evaluate regression model performance.

How do I train a solubility predictor on the Delaney ESOL benchmark?▼

You can train a solubility predictor on the Delaney ESOL benchmark by running executable training scripts with DeepChem loaders, applying scaffold-based splitting, and reporting the resulting test metrics.

Related Skills

Explore More Data & Analytics Skills→
crazymsncrazymsn
22

deepchem

Run end-to-end molecular property prediction workflows with DeepChem.

Community
Advanced
choxoschoxos
9

Model Evaluation Patterns

Quantifies model performance for classification, regression, and survival tasks using yardstick R packages.

Community
Advanced
Lord1EgyptLord1Egypt
2

deepchem

Train graph neural networks for molecular property prediction on chemical datasets.

Community
Advanced
biomaps-infrabiomaps-infra

lib-deepchem

Trains and applies DeepChem machine learning models for molecular property prediction.

Official
Advanced