boltz

Predict protein, nucleic-acid, and small-molecule complex structures with Boltz-2.

288|34|Updated Jul 6, 2026

Search Tags:#science #binding affinity #protein structure prediction #complex structure #AlphaFold alternative

Ready to run

npx skills add https://github.com/PKU-YuanGroup/OpenAI4S --skill boltz-pku-yuangroup

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: boltz
Source: https://github.com/PKU-YuanGroup/OpenAI4S/tree/main/skills/boltz
Command: npx skills add https://github.com/PKU-YuanGroup/OpenAI4S --skill boltz-pku-yuangroup

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires boltz, torch, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill automates structure prediction for protein, nucleic-acid, and small-molecule complexes, providing validation tools and alternative models for research.

Core Features & Use Cases

Structure Prediction: Predict the structure of complex molecules using Boltz-2 and optional binding-affinity prediction.
Validation: Validate designed binders against a target protein.
Co-folding: Co-fold a protein with a ligand.
Alternative to AlphaFold3: Offers an open-source alternative for AlphaFold3 with optional binding-affinity prediction.
Use Case: A researcher needs to validate a binder for a protein, could use this Skill to predict the structure and binding affinity, thereby saving time in traditional experimental methods.

Quick Start

Use the boltz skill to predict the structure of a protein complex from the provided sequences.

Frequently Asked Questions about boltz

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

FAQPage Schema

How do I predict protein complex structures and binding affinity as an AlphaFold alternative?▼

You can predict protein complex structures and binding affinity using Boltz-2, an open-source alternative to AlphaFold3 that utilizes diffusion and MSA models to generate accurate complex structures.

Can I co-fold a protein with a ligand and validate designed binders?▼

Yes, you can co-fold proteins with ligands and validate designed binders against a target protein by predicting both the complex structure and optional binding affinity using the Boltz-2 model.

What is needed to run Boltz-2 for small-molecule and nucleic-acid complex prediction?▼

Running Boltz-2 for small-molecule and nucleic-acid complex prediction requires a GPU for performance, along with the torch and boltz dependencies installed in your environment.

Does Boltz-2 support structure prediction for nucleic acids and small molecules?▼

Yes, Boltz-2 supports structure prediction for nucleic acids and small molecules, allowing researchers to model diverse complexes and validate interactions without relying solely on AlphaFold3.

Why use diffusion and MSA models for complex structure prediction instead of AlphaFold3?▼

Using diffusion and MSA models for complex structure prediction provides an open-source alternative to AlphaFold3, offering optional binding-affinity prediction to validate designed binders and co-fold proteins with ligands.

Are there limitations when predicting complex structures without a GPU?▼

Predicting complex structures without a GPU is a limitation because Boltz-2 requires GPU hardware for performance when utilizing diffusion and MSA models to generate protein, nucleic-acid, and small-molecule complexes.