Babak Emami

We formulate fixed–backbone computational protein design (CPD) as a quadratic Hamiltonian over rotamer variables and benchmark Quantum Computing Inc. 's Dirac-3 entropy-based optimizer against optimal solutions from an exact classical Cost Function Network (CFN) baseline. On seven small proteins (493–943 variables), Dirac-3 attains energies within 0.2 -- 3.7% of CFN. For two larger instances (1RIS, 1GVP; >3,000 variables), we introduce a METIS-based position-graph partitioning with iterative refinement: energies are within 10% of CFN while achieving shorter wall-clock times. We analyze key device hyper-parameters and show that moderate settings preserve energy quality and markedly reduce runtime. Overall, Dirac-3 exhibits gentle runtime scaling and robust performance, indicating practical promise for large CPD instances where exact classical methods become time-prohibitive.