QO-BRA: A Quantum Operator-Based Autoencoder for De Novo Molecular Design

Area of Science:

• Quantum computing
• Computational chemistry
• Bioinformatics

Background:

• De novo molecular design is crucial for discovering novel therapeutics and materials.
• Current generative models face challenges in ensuring the novelty, uniqueness, and validity of generated molecules.
• Quantum computing offers potential advantages for complex computational tasks in chemistry and biology.

Purpose of the Study:

• To introduce QO-BRA (Quantum Operator-Based Real Amplitude autoencoder), a novel variational quantum autoencoder for de novo molecular design.
• To leverage quantum circuits for enhanced molecular representation and generative processes.
• To demonstrate the application of QO-BRA in designing specific metalloproteins.

Main Methods:

• Utilized quantum circuits for real-amplitude encoding.
• Employed the SWAP test for error estimation during back-propagation.
• Implemented adjoint encoder and decoder operators for unitary transformations.

Main Results:

• QO-BRA ensures accurate reconstruction of molecular structures.
• The generative process guarantees novelty, uniqueness, and validity of generated samples.
• Successfully applied QO-BRA to design Ca2+, Mg2+, and Zn2+-binding metalloproteins.

Conclusions:

• QO-BRA represents a significant advancement in quantum-enhanced de novo molecular design.
• The model demonstrates efficacy even with modest datasets.
• This approach holds promise for accelerating the discovery of functional biomolecules.