Quantum Protein Folding with VQE

Overview

This project demonstrates how quantum algorithms on quantum simulators can tackle protein folding by treating it as an energy minimization problem on a rugged landscape—a task where quantum exploration of state spaces offers advantages over classical optimization methods when no structural templates exist.

Using a simplified HP (Hydrophobic-Polar) model, we apply Variational Quantum Eigensolver (VQE) to find low-energy conformations, demonstrating quantum computing's potential for de novo protein structure prediction.

Motivation

Current protein structure prediction methods face challenges:

• AlphaFold excels when structural homologs exist but struggles with novel sequences
• Physics-based methods (molecular dynamics, Monte Carlo) are computationally expensive for complex energy landscapes
• Quantum computing offers potential advantages in exploring the exponentially large conformational space

This project targets scenarios where classical methods fall short: sequences lacking structural templates, requiring true physics-based energy minimization.

Quantum Advantage

• State space exploration: Quantum superposition allows parallel exploration of conformational states
• Energy landscape navigation: Quantum tunneling can escape local minima that trap classical optimizers
• Scalability: VQE provides a NISQ-era approach viable on current quantum hardware

Project Structure

quantum-protein-folding/
├── README.md                    # This file
├── FRONTEND_LAUNCH.md           # Quick start guide for frontend
├── requirements.txt             # Python dependencies
├── backend/
│   ├── src/
│   │   ├── hp_model.py          # HP protein model implementation
│   │   ├── hamiltonian.py       # Energy Hamiltonian construction
│   │   ├── vqe_optimizer.py     # VQE-based optimization
│   │   ├── visualization.py     # Results visualization
│   │   └── utils.py             # Helper functions
│   └── examples/
│       ├── quick_start.py       # CLI demo
│       ├── test_small.py        # Quick test
│       └── demo_notebook.ipynb  # Interactive notebook
├── frontend/
│   ├── app.py                   # Streamlit web application
│   ├── requirements.txt         # Frontend dependencies
│   └── README.md                # Frontend documentation
├── results/                     # Experimental results
└── docs/                        # Additional documentation

Qiskit Components Used

• qiskit: QuantumCircuit, quantum gates
• qiskit.quantum_info: SparsePauliOp for Hamiltonian representation
• qiskit_algorithms: VQE, optimizers (COBYLA, SLSQP)
• qiskit.circuit.library: TwoLocal, EfficientSU2 ansätze
• qiskit.primitives: Estimator for expectation values
• qiskit_aer: AerSimulator for quantum simulation
• qiskit.visualization: Circuit and results visualization

Installation

pip install -r requirements.txt

Quick Start

Option 1: Web Interface (Recommended)

cd frontend
streamlit run app.py

Opens interactive web application at http://localhost:8501

Option 2: Command Line

cd backend/examples
python quick_start.py

Option 3: Python API

from backend.src.hp_model import HPModel
from backend.src.vqe_optimizer import VQEProteinFolder

# Define a simple HP sequence
sequence = "HPHPPHHPHH"

# Create model and run VQE
model = HPModel(sequence)
optimizer = VQEProteinFolder(model)
result = optimizer.optimize()

print(f"Lowest energy: {result.eigenvalue}")
print(f"Best conformation: {result.conformation}")

Option 4: Jupyter Notebook

cd backend/examples
jupyter notebook demo_notebook.ipynb

Scientific Scope

This is a toy but scientifically meaningful prototype due to:

• Current quantum hardware limitations (noise, qubit count)
• Simplified HP model vs. full atomic physics
• Focus on proof-of-concept for quantum advantage

However, it demonstrates core principles applicable to real quantum protein folding once hardware scales.

Comparison Baseline

We compare against:

• Classical VQE simulation: Same algorithm, classical computer
• Random search: Baseline optimization
• Greedy optimization: Local search methods
• (Future) Molecular dynamics simulation for small sequences

Future Directions

Near-term Extensions

• Extend to 3D cubic lattice (6 directions, 3 qubits/move)
• Implement Q-score and contact order validation
• Test on HP benchmark sequences
• Add radius of gyration analysis

Mid-term Goals

• FCC lattice for more realistic geometry (12 directions)
• Incorporate more amino acid types beyond H/P
• Full atomic reconstruction with side chains
• RMSD validation against known structures

Long-term Vision

• Incorporate realistic force fields (AMBER, CHARMM)
• Hybrid quantum-classical pipeline (coarse-grained → all-atom)
• Error mitigation for real quantum hardware
• Integration with existing tools (AlphaFold, Rosetta)

See detailed discussions in:

• LATTICE_MODELS.md - Lattice selection and trade-offs
• VALIDATION_METRICS.md - Validation approaches and metrics

References

• Perdomo-Ortiz et al., "Finding low-energy conformations of lattice protein models by quantum annealing" (2012)
• Robert et al., "Resource-efficient quantum algorithm for protein folding" (2019)
• Qiskit documentation: https://qiskit.org/documentation/

License

MIT License

Authors

IBM Qiskit Fall Fest Hackathon Project