Protein Structure Prediction Algorithm Notebooks

A comprehensive educational resource for understanding the core algorithms of modern protein structure prediction models, including AlphaFold2, AlphaFold3, Boltz-1, and Boltz-2.

Overview

This repository provides detailed Jupyter notebooks that explain the key algorithms from each model's architecture, with:

• Pseudocode explanations from original papers
• NumPy implementations for educational purposes
• Source code references to official implementations
• Test examples to verify understanding

Repository Structure

alphafold-notebooks/
├── alphafold2/              # AlphaFold2 (32 algorithms)
│   ├── notebooks/           # Algorithm explanation notebooks
│   ├── source/              # Source code reference (local copy)
│   ├── ref-src/             # External reference repositories (submodules)
│   ├── references/          # Related papers info
│   ├── presentations/       # Slides and presentations
│   ├── applications/        # Application examples
│   └── AF2REFPAPERS.md      # 83 reference papers
│
├── alphafold3/              # AlphaFold3 (23 algorithms)
│   ├── notebooks/
│   ├── ref-src/             # External reference repositories (submodules)
│   └── AF3REFPAPERS.md      # 50 reference papers
│
├── boltz/                   # Boltz-1 (20 algorithms)
│   ├── notebooks/
│   ├── ref-src/             # External reference repositories (submodules)
│   └── BOLTZREFPAPERS.md    # 46 reference papers
│
├── boltz2/                  # Boltz-2 (10 new algorithms)
│   ├── notebooks/
│   └── BOLTZ2REFPAPERS.md   # 50 reference papers
│
├── finetuning/              # Fine-tuning framework
│   ├── configs/             # Task configs (25+ types)
│   ├── modules/             # LoRA, Adapter, Prompt Tuning
│   ├── heads/               # 15+ specialized prediction heads
│   ├── trainers/            # Training with DDP, AMP
│   ├── data/                # 10+ dataset classes
│   └── examples/            # Tutorial notebooks
│
└── assets/
    └── images/              # Shared image resources

Model Comparison

Model	Key Architecture	Algorithms Covered
AlphaFold2	Evoformer + IPA Structure Module	32
AlphaFold3	MSA Module + Pairformer + Diffusion	23
Boltz-1	Pairformer + Diffusion (open source)	20
Boltz-2	+ Affinity Prediction (binding affinity)	10 (new)

Reference Source Code (Git Submodules)

AlphaFold2 References

Repository	Description	URL
alphafold-official	DeepMind's official AlphaFold2	deepmind/alphafold
openfold	Trainable PyTorch reproduction	aqlaboratory/openfold
colabfold	Fast AlphaFold on Google Colab	sokrypton/ColabFold
mmseqs2	Fast sequence search tool	soedinglab/MMseqs2
hh-suite	HMM-based sequence search	soedinglab/hh-suite
trRosetta2	Alternative structure prediction	RosettaCommons/trRosetta2
esm	Meta's protein language models	facebookresearch/esm
unirep	UniRep protein representation	churchlab/UniRep
seqvec	ELMo for proteins	rostlab/SeqVec

AlphaFold3 References

Repository	Description	URL
alphafold3-official	DeepMind's official AlphaFold3	google-deepmind/alphafold3
alphafold3-pytorch	PyTorch reproduction by lucidrains	lucidrains/alphafold3-pytorch
alphafold3-walkthrough	Architecture walkthrough	shenyichong/alphafold3-architecture-walkthrough

Boltz References

Repository	Description	URL
boltz-official	Official Boltz-1 & Boltz-2	jwohlwend/boltz
boltzina	Boltz for virtual screening	ohuelab/boltzina

Clone with Submodules

# Clone with all submodules
git clone --recursive https://github.com/your-repo/alphafold-notebooks.git

# Or initialize submodules after clone
git submodule update --init --recursive

Quick Start

Each model directory contains:

• ALGORITHM_INDEX.md - Complete algorithm listing with notebooks and source references
• *REFPAPERS.md - Comprehensive reference paper list

Algorithm Notebooks

Model	Algorithms	Index
AlphaFold2	32 (Evoformer, IPA, FAPE...)	Index
AlphaFold3	23 (Diffusion, Pairformer...)	Index
Boltz-1	20 (AtomEncoder, Confidence...)	Index
Boltz-2	10 (Affinity, Contact Cond...)	Index

Key Topics Covered

Representation Learning

• MSA Processing (Row/Column Attention)
• Outer Product Mean
• Triangle Multiplication & Attention
• Pairformer Stack

Structure Prediction

• Invariant Point Attention (IPA) - AlphaFold2
• Diffusion Transformer - AlphaFold3/Boltz
• Atom Cross Attention

Confidence & Loss

• pLDDT, pAE, pTM metrics
• FAPE Loss
• Diffusion Loss
• Binding Affinity (Boltz-2)

🔧 Fine-tuning Framework (NEW!)

We provide a comprehensive fine-tuning framework for adapting protein structure prediction models to downstream tasks.

👉 Full Fine-tuning Guide

Supported Models

Model	Framework	Fine-tuning Support
AlphaFold2	JAX/Haiku	✅ Full, Head-only, LoRA
AlphaFold3	JAX/Haiku	✅ Full, Head-only, LoRA
Boltz-1	PyTorch	✅ Full, LoRA, Adapter
Boltz-2	PyTorch	✅ Full, LoRA, Adapter

Supported Tasks (50+ Task Types)

💊 Drug Discovery

Task	Outputs	Applications
Binding Affinity	pKd, pIC50, ΔG, Ki	Lead optimization, SAR
Virtual Screening	Hit probability, ranking	HTS prioritization
ADMET	Absorption, metabolism, toxicity	Compound triage

🔬 Protein Engineering

Task	Outputs	Applications
Stability	ΔΔG, Tm shift	Thermostabilization
Solubility	Expression score	Biomanufacturing
Mutation Effects	Fitness, pathogenicity	Variant analysis

🧫 Antibody Design

Task	Outputs	Applications
Affinity Maturation	CDR binding, ΔΔG	Therapeutic optimization
Humanization	Humanness score	Drug development
Developability	Aggregation, viscosity	Manufacturing

⚗️ Enzyme Engineering

Task	Outputs	Applications
Activity	kcat, Km, kcat/Km	Catalyst design
Specificity	Substrate profiles	Industrial enzymes
Directed Evolution	Fitness landscapes	Protein engineering

🔗 Protein-Protein Interactions

Task	Outputs	Applications
PPI Binding	Kd, interface stability	Complex analysis
Interface Prediction	Contact residues	Structure analysis
Hot Spot Detection	ΔΔG per residue	PPI drug targets

🧬 Function & Immunology

Task	Outputs	Applications
GO Terms	MF, BP, CC	Annotation
B-cell Epitopes	Epitope probability	Vaccine design
T-cell Epitopes	MHC binding	Immunotherapy

Quick Start

from finetuning import TaskRegistry, create_finetuning_pipeline
from finetuning.modules import LoRAModule

# List all 50+ tasks
print(TaskRegistry.list_all_tasks())

# Get task recommendations
info = TaskRegistry.get_task_info("binding_affinity")
print(f"Recommended LoRA rank: {info.recommended_rank}")

# Create pipeline
pipeline = create_finetuning_pipeline(
    task="binding_affinity",
    base_model=model,
    strategy="lora",
)

---

References

Core Papers

Model	Paper	DOI
AlphaFold2	Highly accurate protein structure prediction with AlphaFold	Nature 2021
AlphaFold3	Accurate structure prediction of biomolecular interactions	Nature 2024
Boltz-1	Democratizing Biomolecular Interaction Modeling	bioRxiv 2024
Boltz-2	Towards Accurate and Efficient Binding Affinity Prediction	bioRxiv 2025

Full Reference Lists

Each model has a comprehensive reference paper list covering foundational works, methods, and applications:

• AlphaFold2 References - 83 papers (structural biology, deep learning, MSA)
• AlphaFold3 References - 50 papers (diffusion models, transformers, confidence)
• Boltz-1 References - 46 papers (open source, docking, PLMs)
• Boltz-2 References - 50 papers (binding affinity, FEP, virtual screening)

License

Educational use only. Please refer to the original papers and repositories for licensing information.