Active Space Selection Benchmarks Accelerate Quantum Drug Discovery with VQE for Lovastatin and More

Quantum computing holds immense potential for revolutionising drug discovery, yet current limitations in quantum hardware hinder its application to complex molecules. Zhi Yin, Xiaoran Li, and Zhupeng Han, from QureGenAI and Ningbo University of Technology, alongside Shengyu Zhang from Tencent Quantum Lab and colleagues, address this challenge by systematically evaluating a crucial technique called active space selection. Their work introduces a comprehensive benchmark that assesses how effectively this method prepares molecules for quantum calculations within the variational quantum eigensolver (VQE) pipeline. By testing representative drug-like molecules, such as lovastatin and morphine, and comparing results from both simulated and real quantum processors, the team demonstrates the significant impact of active space choices on both the accuracy and efficiency of quantum drug discovery, establishing a foundation for future advances in hardware and algorithm development.

Their work introduces a comprehensive benchmark that assesses how effectively this method balances accuracy and efficiency.

Benchmarking VQE for Drug-Like Molecules

The team tested the technique on representative drug-like molecules, such as lovastatin, oseltamivir, and morphine, and compared results from both simulated and real quantum processors. This dual approach allows for a robust comparison of performance across different algorithmic strategies and hardware platforms. The study goes beyond simple energy calculations, incorporating both chemistry-centric and architecture-centric metrics to provide a holistic evaluation.

Impact of Active Space Choices

Results demonstrate the significant impact of active space choices on both the accuracy and efficiency of quantum drug discovery. Researchers identified correlations between molecular characteristics and successful VQE performance under real-world noise conditions. The work establishes the first systematic benchmark for active space driven VQE, laying the groundwork for future hardware-algorithm co-design studies.

Evaluating VQE Performance

Scientists have established a systematic benchmark for evaluating VQE algorithms, specifically focusing on active space selection. Evaluations employed both classical simulations and execution on quantum processing units (QPUs), utilizing the unitary coupled-cluster with singles and doubles (UCCSD) and hardware-efficient ansatz (HEA) methods. This comprehensive approach reveals critical trade-offs between achieving chemical fidelity and maintaining feasibility on current hardware.

Scaling and Convergence

Experiments on superconducting quantum processors identified that while larger processors can achieve comparable final energies, they exhibit increased convergence oscillations, suggesting that current VQE applications may be limited by circuit depth rather than qubit count. Researchers found that VQE achieves robust convergence on current quantum hardware for smaller active spaces, successfully optimizing calculations within a reasonable number of iterations.

Future Directions

The team openly releases the benchmark suite, including molecular geometries, active space configurations, classical reference data, and VQE evaluation scripts, to facilitate reproducibility and further community-driven extensions. Researchers propose developing adaptive active space protocols, designing algorithms tailored to active space characteristics, and employing hierarchical active space refinement techniques to maximize the potential of quantum computing in drug discovery.