Exploring Chemical Space with Chemistry-Inspired Dynamic Quantum Circuits in the NISQ Era

Area of Science:

• Computational Chemistry
• Quantum Computing
• Artificial Intelligence

Background:

• Molecular structure generation is vital for drug design and materials science.
• Classical generative models face limitations in data, computational resources, and parameter efficiency.
• Quantum computing offers potential for reduced computational overhead via quantum parallelism and entanglement.

Purpose of the Study:

• To design and evaluate a quantum-based molecular generator (QMG) for small molecules (C, N, O).
• To demonstrate the efficiency and validity of quantum-generated molecular structures.
• To explore the QMG's capability for targeted chemical space exploration.

Main Methods:

• Developed a QMG utilizing quantum circuits with constrained output states.
• Employed quantum superposition and entanglement for efficient molecular enumeration.
• Used Bayesian optimization to validate and refine generated molecular structures.

Main Results:

• The QMG, with only 134 parameters, enumerated all structures up to nine heavy atoms.
• Generated molecules showed high validity and uniqueness, comparable to classical models.
• The QMG successfully generated molecules with specific functional groups by fixing circuit parameters.

Conclusions:

• The compact QMG design significantly reduces parameter demands for molecular generation.
• Quantum-based models offer an efficient advantage for exploring chemical space.
• The QMG shows promise for targeted applications, particularly in drug discovery.