alphafold3

What problem does it solve?

This Skill enables automated joint prediction of biomolecular structures for proteins, DNA/RNA, ligands, ions, and their complexes, reducing manual modeling time and enabling rapid hypothesis testing in structural biology.

Core Features & Use Cases

• Diffusion-based, high-accuracy structure prediction for biomolecular assemblies including protein complexes, protein-DNA/RNA complexes, and protein-ligand interactions.
• Supports multiple entity types (protein, DNA, RNA, ligand, ion) and complex formation scenarios for research and discovery workflows.
• Useful in drug discovery, protein engineering, and mechanistic studies of biomolecular interactions.

Quick Start

Run AlphaFold3 in a GPU-enabled Docker environment using the official image, mounting model directories, genetic databases, and input/output folders. Example: docker run --rm --gpus all
-v /path/to/models:/root/models
-v /path/to/databases:/root/public_databases
-v $(pwd)/input:/root/af_input
-v $(pwd)/output:/root/af_output
ghcr.io/open-prophetdb/alphafold3:arm64-blackwell
python run_alphafold.py --json_path=/root/af_input/fold_input.json --model_dir=/root/models --output_dir=/root/af_output