HIV-1 protease cleavage sites detection with a quantum convolutional neural network algorithm

Area of Science:

• Quantum Computing
• Bioinformatics
• Computational Biology

Background:

• Human immunodeficiency virus type 1 (HIV-1) protease is vital for viral maturation, making cleavage site prediction crucial for therapeutic development.
• Accurate prediction of HIV-1 protease cleavage sites aids in understanding viral pathogenesis and designing inhibitors.

Purpose of the Study:

• To develop a quantum convolutional neural network (QCNN) framework integrated with neural quantum embedding (NQE) for predicting HIV-1 protease cleavage sites.
• To enhance feature representation in quantum space for improved sequence classification accuracy.

Main Methods:

• A QCNN-based framework with NQE was designed to predict HIV-1 protease cleavage sites from amino acid sequences.
• The quantum model was evaluated against classical neural networks using four HIV-1 protease cleavage site datasets under simulated noisy and noiseless quantum environments.
• Experiments explored scalability and parameter efficiency using varying numbers of qubits and trainable parameters.

Main Results:

• QCNN models with NQE (angle and amplitude encoding) achieved higher classification accuracy than classical neural networks, with average accuracies of 0.9146 (4-qubit) and 0.8929 (8-qubit).
• The QCNN with NQE (ZZ feature map and angle encoding) demonstrated stable performance under simulated quantum hardware noise.
• The quantum approach outperformed classical counterparts, even under simulated noise conditions.

Conclusions:

• This study presents the first NQE-augmented QCNN for HIV-1 protease cleavage site prediction, demonstrating quantum machine learning's potential in biomedical sequence analysis.
• NQE-enhanced QCNNs show promise for scalable, noise-resilient quantum machine learning in bioinformatics.
• The findings lay the groundwork for future quantum-based bioinformatics analyses and therapeutic inhibitor development.