boltz

Predict protein complex structures with Boltz-1/Boltz-2 and output CIF models.

151|20|Updated Jan 15, 2026
One-click install
npx skills add https://github.com/adaptyvbio/protein-design-skills --skill boltz
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: boltz
Source: https://github.com/adaptyvbio/protein-design-skills/tree/main/skills/boltz
Command: npx skills add https://github.com/adaptyvbio/protein-design-skills --skill boltz

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

Boltz provides an open-source, GPU-friendly pathway for predicting protein complex structures using Boltz-1 and Boltz-2, filling gaps left by closed tools and enabling local experimentation.

Core Features & Use Cases

  • Open-source structure prediction for protein complexes with Boltz-1/Boltz-2.
  • Flexible deployment via Modal or local Python, with guidance for installation.
  • Outputs include CIF models, confidence metrics (ptm, iptm, pLDDT) and PAE matrices.
  • Use cases: validating designed binders; exploring protein-ligand complexes; benchmarking structure prediction in local HPC.

Quick Start

Run Boltz to predict a complex: modal run modal_boltz.py --input-faa complex.fasta --out-dir predictions/

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 using an open-source model?

You can predict protein complex structures by running Boltz-1 or Boltz-2 via a Python workflow. It processes an input FASTA file and outputs CIF models along with confidence metrics and PAE matrices.

What is needed to run Boltz for local protein structure prediction?

Running Boltz locally requires Python 3.10 or higher and CUDA-enabled GPUs. The workflow supports local installations to predict complexes and output CIF models without relying on cloud resources.

Can I use Modal to run GPU-based protein structure prediction?

Yes, Boltz supports Modal-based deployment to run structure prediction on remote GPUs. You execute a Python script with your input FASTA file and output directory to generate predicted complexes.

What outputs do I get when validating designed protein binders with Boltz?

Validating designed binders with Boltz outputs CIF structural models, confidence metrics including ptm, iptm, and pLDDT scores, and PAE matrices to help assess prediction quality.

Does Boltz work for predicting protein-ligand complexes?

Yes, Boltz supports predicting protein-ligand complexes using its open-source models. It generates CIF models and confidence metrics for the predicted interactions.

Why use Boltz instead of AlphaFold2 for protein structure prediction?

Boltz provides an open-source alternative for protein complex prediction when local GPU resources are preferred or closed tools are unavailable. It runs via Modal or local Python environments to generate comparable structural outputs.