openfold3

Predict 3D structures of proteins, nucleic acids, and ligands with OpenFold3.

288|34|Updated Jul 6, 2026
One-click install
npx skills add https://github.com/PKU-YuanGroup/OpenAI4S --skill openfold3-pku-yuangroup
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: openfold3
Source: https://github.com/PKU-YuanGroup/OpenAI4S/tree/main/skills/openfold3
Command: npx skills add https://github.com/PKU-YuanGroup/OpenAI4S --skill openfold3-pku-yuangroup

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires openfold3, deepspeed, cutlass, huggingface, nvidia-cutlass, boto3, awscrt, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill solves the problem of accurately predicting complex structures of proteins, nucleic acids, and ligands by leveraging OpenFold3, an open-source reproduction of AlphaFold3.

Core Features & Use Cases

  • Protein Structure Prediction: Predict the 3D structure of proteins with high accuracy.
  • Nucleic Acid and Ligand Structures: Also capable of predicting the structures of nucleic acids and ligands.
  • Open-Source: Uses an Apache-2.0 licensed implementation of AlphaFold3.
  • Use Case: For scientists who need to quickly predict the structure of a new protein or ligand.

Quick Start

Run the OpenFold3 skill with the following command:

run_openfold predict --query_json=queries.json --output-dir out/

Frequently Asked Questions about openfold3

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

FAQPage Schema
How do I predict protein and ligand complex structures using an open-source AlphaFold3 implementation?

You can predict protein and ligand complex structures using OpenFold3, an open-source reproduction of AlphaFold3. It accurately models 3D structures of proteins, nucleic acids, and ligands via JSON orchestration and Python execution in persistent kernels.

What's the best way to run protein structure prediction for bioinformatics research?

The best way to run protein structure prediction is using the OpenFold3 skill with a JSON query file. Execute `run_openfold predict --query_json=queries.json --output-dir out/` to generate 3D structural outputs for bioinformatics research.

Do I need a GPU to predict nucleic acid and ligand structures with OpenFold3?

Yes, you need a GPU to process nucleic acid and ligand structure predictions with OpenFold3. GPU acceleration is required to handle the complex computations involved in accurately modeling these 3D molecular structures.

Can I use OpenFold3 for drug discovery and nucleic acid structure prediction?

Yes, you can use OpenFold3 for drug discovery and nucleic acid structure prediction. It is an Apache-2.0 licensed implementation designed to accurately predict complex 3D structures of proteins, nucleic acids, and ligands.

What dependencies are required for OpenFold3 protein structure prediction?

OpenFold3 protein structure prediction requires dependencies including deepspeed, cutlass, huggingface, nvidia-cutlass, boto3, and awscrt. These libraries support the deep learning and data orchestration needed for structural modeling.

Are there limitations when using an open-source AlphaFold3 implementation for complex structures?

OpenFold3 accurately predicts complex structures of proteins, nucleic acids, and ligands but requires significant GPU resources. As an open-source implementation of AlphaFold3, its accuracy depends on the provided JSON query inputs and persistent kernel execution.