AlphaFold 3 for Predicting Biomolecular Structures and Interactions

Author: Xiping Gong
Department of Crop and Soil Sciences, College of Agricultural and Environmental Sciences, University of Georgia, Griffin, GA, USA
📧 xipinggong@uga.edu
🗓️ First Draft: 11/11/2025

---

📘 Overview

This Jupyter Notebook (alphafold3.ipynb) serves as a hands-on tutorial for MIBO 8110 students to learn how to use AlphaFold 3 (AF3) for predicting biomolecular structures and interactions.

It provides a step-by-step workflow showing how AlphaFold 3 integrates protein and small-molecule components into a unified structure prediction system. Students will gain practical experience in using computational structure prediction tools, performing post-analysis, and interpreting results.

---

🧩 Learning Objectives

By working through this notebook, students will learn how to:

• Predict protein 3D structures from amino acid sequences using AlphaFold 3.
• Predict protein–ligand complexes and explore binding interactions.
• Analyze and visualize predicted structures using MDTraj and VMD.
• Calculate RMSD values to assess prediction accuracy.
• Perform post-processing (alignment, file conversion, and interaction mapping).

---

⚙️ Prerequisites and Setup

• AlphaFold 3 must be installed.
  Official GitHub: https://github.com/google-deepmind/alphafold3

• Environment:
  AF3 is already installed on the GACRC teaching node.
  Log in using your UGA MyID:

$ ssh <MyID>@teach.gacrc.uga.edu

• Dependencies:
  Some useful Python packages like mdtraj, openmm, and biopython are preinstalled.
  For additional setup help, see: https://xipinggong.com/files/tutorials/setup.html

---

🧪 Sections Overview

1. Protein Structure Prediction

Students learn how to:

1. Download a PDB structure (e.g., human serum albumin 7AAI).
2. Extract the protein chain (extract_pdb.py).
3. Prepare input JSON for AF3 (get_json_for_af3.py).
4. Run AF3 via SLURM (sbatch af3.sh 7AAI_protein.json).
5. Convert and align output CIF → PDB → aligned structure (cif2pdb.py, align_pdb.py).
6. Visualize using VMD.

---

2. Protein–Ligand Interaction Prediction

• Extend the workflow to include ligands (e.g., 8PF and MYR in the 7AAI system).
• Create a multi-component JSON input specifying both protein and ligands.
• Run AF3 again and align the predicted complexes.
• Discuss accuracy and limitations (e.g., training set overlap).

---

3. Structural Interaction Analysis

• Use the provided scripts to identify binding residues
• Calculate the RMSD values
• Example systems: PFOA–human serum albumin (PDB: 7AAI) & MYR–human serum albumin interactions