Speaker
Description
Protein folding is an NP-hard problem, making accurate prediction of stable three-dimensional structures computationally demanding. Despite advances in classical methods, an efficient general solution remains unresolved. Quantum computing has improved the study of small, simplified folding models, though scaling to realistic proteins is still a challenge. Here, we propose a novel turn-based encoding for protein folding on a cubic lattice that incorporates solvent interaction energies to capture both structural and energetic properties and that reduces locality to three. In contrast to previous formulations, we significantly decrease circuit complexity. We compare three ansatze- EfficientSU2, RealAmplitude, and TwoLocal for eight peptides on AerSimulator, and deploy the best-performing ansatz within CVaR-VQE on IBM Brisbane. Quantum optimization is performed via VQE on hardware and simulation, while IBM CPLEX and simulated annealing serve as classical benchmarks. Results show improved performance on quantum hardware compared to simulation, highlighting the potential of quantum systems over classical methods. RMSD analysis against Protein Data Bank reference structures further confirms that the proposed encoding better captures realistic protein folds. Overall, this study establishes a foundation for quantum-enhanced protein folding simulations and paves the way for broader applications in computational biophysics.
