Protein Structure Prediction with Quantum Algorithms (Part 2/2)

Predicting protein structures from amino acid sequences is a key challenge in computational biology with implications for drug discovery. Quantum approaches using coarse-grained lattice models and variational algorithms have been explored, but often require explicit Hamiltonian construction, ancillary qubits, and deep circuits. We present a scalable quantum workflow that bypasses Hamiltonian encoding by employing a problem-agnostic ansatz trained to minimize a classically computable energy-based cost function. This reduces circuit depth and enables inclusion of higher-order interactions. We benchmark a hardware-efficient ansatz on proteins up to 26 amino acids modeled on tetrahedral, body-centered cubic, and face-centered cubic lattices with second-nearest-neighbor interactions. Performance is evaluated on a noise-free simulator and the ibm_kingston quantum device using multiple quality metrics. Our results demonstrate improved scalability and flexibility compared to prior methods, while highlighting challenges and opportunities for advancing quantum protein structure prediction.