molfeat

Name: molfeat
Availability: InStock
Author: rubensliv

Convert chemical structures into machine-learning-ready feature vectors for ML workflows.

15|2|Updated Dec 17, 2025

Search Tags:#featurization #virtual-screening #ml-pipelines #molecular #qsar #pretrained-embeddings

Ready to run

npx skills add https://github.com/rubensliv/k-dense-ai --skill molfeat-rubensliv

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: molfeat
Source: https://github.com/rubensliv/k-dense-ai/tree/main/scientific-skills/molfeat
Command: npx skills add https://github.com/rubensliv/k-dense-ai --skill molfeat-rubensliv

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Molfeat provides a unified suite of molecular featurization tools to convert chemical structures into machine-learning-ready feature vectors, enabling QSAR modeling, virtual screening, and similarity analysis.

Core Features & Use Cases

100+ featurizers and pretrained embeddings covering fingerprints, descriptors, pharmacophore and shape descriptors, and graph representations.
Scikit-learn compatible transformers, parallel processing, caching, and robust pipelines for end-to-end ML workflows across chemistry domains.
Use cases include QSAR modeling, virtual screening, similarity searching, clustering, interpretability, and model deployment.

Quick Start

Create a MoleculeTransformer with an FPCalculator (e.g., FPCalculator('ecfp')) and apply it to your SMILES list to obtain feature matrices.

Frequently Asked Questions about molfeat

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

FAQPage Schema

How do I convert SMILES to molecular feature vectors for QSAR modeling?▼

Molecular featurization converts SMILES strings into ML-ready feature vectors using FPCalculator and MoleculeTransformer, enabling QSAR modeling. You create a transformer with a specified featurizer and apply it directly to your SMILES list to obtain numpy or pytorch matrices.

What types of molecular featurizers are available for virtual screening?▼

Molecular featurizers for virtual screening include traditional fingerprints, descriptor-based features, pharmacophore and shape descriptors, pretrained transformer embeddings, and graph representations. Over 100 featurizers and pretrained embeddings are available to cover diverse chemistry domains.

Can I use scikit-learn compatible transformers for molecular featurization in ML pipelines?▼

Yes, molecular featurization provides scikit-learn compatible transformers that support parallel processing and caching. These transformers output numpy or pytorch arrays, allowing seamless integration into end-to-end ML pipelines for clustering, similarity searching, and model deployment.

Does molfeat support pretrained embeddings for molecular similarity search?▼

Yes, pretrained transformer embeddings are supported for molecular similarity search and QSAR modeling. The framework provides calculators and model store components to generate these embeddings, allowing you to compute structural similarities using advanced learned representations.

What is the best way to scale molecular featurization for large compound libraries?▼

To scale molecular featurization for large compound libraries, use the built-in parallel processing and caching capabilities of the scikit-learn compatible transformers. This approach accelerates feature vector generation while avoiding redundant computations across your screening workflows.

Are there optional dependencies for graph representations in molecular featurization?▼

Yes, molecular featurization supports optional dependencies specifically for graph representations and pretrained transformer embeddings. This modular design ensures you only install the required packages for your specific QSAR or similarity analysis workflow.