QuantumHoneycombPSP

QuantumHoneycombPSP is a Python-based project designed to model protein structure prediction (PSP) using quantum computing techniques. The project includes various modules to encode sequences, generate energy matrices, and create energy functions for different lattice types. It also supports the conversion of these energy functions into Quadratic Unconstrained Binary Optimization (QUBO) and Ising models. Then runs the QUBO model on D-Wave's quantum annealer to predict the protein structure.

Features

• Sequence Encoding: Encode protein sequences using different models such as HP, HPAB, and WHPAB.
• Energy Matrix Generation: Generate energy matrices based on the encoded sequences and selected energy models.
• Energy Function Creation: Create energy functions for different lattice types (4, 6, 8, 12) and convert them into QUBO and Ising models.
• Interaction Calculation: Calculate interactions between amino acids in the sequence and generate corresponding energy values.
• QUBO Variable Mapping: Map QUBO variables to a new format for further processing.
• Running QUBO on D-Wave: Run the QUBO model on D-Wave's quantum annealer to predict the protein structure.
• Getting Sample Dates: Get sample data from D-Wave's quantum annealer to predict the protein structure.
• Structure Imager: Generate images of protein structures based on the encoded sequences.

Related Works

• Quantum Computing in Protein Folding: Integrating Lattice Models and Energy Functions is a Master's Thesis. Here is the abstract: Protein folding is an NP-Hard problem in computational biology due to the three-dimensional nature of proteins and the vast conformational space. Quantum computing shows promise in addressing this challenge by using unique quantum algorithms to explore protein folding landscapes more efficiently. Key contributions of this thesis include: the Relative Normalized Movement Score, a novel metric for evaluating protein structure fidelity in a distance-independent manner, novel energy models, such as the Hydrophobic-Polar-Acidic-Basic (HPAB) model, and the development and application of CA-2-HCOMB, a Protein Chain Lattice Fitting (PCLF) program, which simplifies protein structures by fitting alpha-carbon traces into discrete lattice models to reduce computational complexity while maintaining structural fidelity. Building upon previous research, we formulated a novel energy encoding for both the Triangular Prismatic Honeycomb (HCOMB-8) lattice model and the Tetrahedral-Octahedral Honeycomb (HCOMB-12) lattice model, also known as the face-centered cubic (FCC) lattice. These lattice models enable more precise protein structure modeling and can be reduced to a Quadratic Unconstrained Binary Optimization (QUBO) problem. To demonstrate the proof of concept, we folded a small peptide sequence using the HCOMB-8 and HCOMB-12 models with the LeapHybrid Algorithm on D-Wave's quantum computer.

• CA-2-HCOMB is a GitHub Repository. CA-2-HCOMB is a Python tool for simplifying single-chain protein structures from PDB files into manageable models like various honeycombs, retaining essential information for in-depth computations. It's efficient for large datasets and machine learning, and user-friendly for research and education.

Installation

To install the required dependencies, run:

pip install -r requirements.txt

Also change the API token in the Annealer.py file to run the QUBO model on D-Wave's quantum annealer.

Usage

QUBO CLI

The main.py script can be used as a command-line interface (CLI) to specify the sequence, energy model, lattice type, and whether to use the binary model. Command-Line Arguments

• sequence: Protein sequence (e.g., 'GAAGA')
• energy_model: Energy model (choices: 'HP', 'HPAB', 'WHPAB', 'MJ')
• lattice_type: Lattice type (choices: 4, 6, 8, 12)
• --binary: Use the binary model (optional flag)

Output

The script will output the following information:

• Encoded sequence
• Interaction matrix
• QUBO, BQM, and Ising models (if --binary is not used)
• Mapped QUBO variables
• Final QUBO equation

Example Command

python main.py GAAGA HP 4 --binary

Visualization CLI

The Visualize.py script can be used to visualize the protein sequence on a lattice. Command-Line Arguments

sequence: Protein sequence (e.g., 'GAAGA') energy_model: Energy model (choices: 'HP', 'HPAB', 'WHPAB', 'MJ') lattice_type: Lattice type (choices: 4, 6, 8, 12) binary_output: Binary move output from QUBO

Example Command

python Visualize.py GAAGA HP 4 101110

Output

The script will output the following information:

• A Matplotlib image of the protein sequence on the lattice
• The energy value of the protein sequence

Example Experiment

We consider the pentapeptide YGGFM, which is Met-enkephalin Tyr-Gly-Gly-Phe-Met an endogenous opioid peptide involved in regulating nociception (pain sensation) in the body. We use ten binary turn qubits to represent the folding directions and 68 ancillary qubits to reduce higher-order terms in the QUBO formulation. We used MJ energy model on a HCOMB12 Lattice.

python main.py YGGFM MJ 12

Then, the top two most occurring proteins, bitstrings, were visualized to produce the following images.

python Visualize.py YGGFM MJ 12 1101001010
python Visualize.py YGGFM MJ 12 1101011010

License

This project is licensed under the MIT License - see the LICENSE file for details.