Table 3 No protein MJ interaction and RMSD values for QUBO approach vs. CP approach. The CP model is anchored to the same starting point as QUBO method to enable fair comparison. The CP results include runs with interaction radius limits set to one grid step (CP 1-NN) and the default interaction radius of 6.5Å (CP). \(t_\textrm{QUBO}\) is the time used to solve a given problem instance (excluding HPO). For all results, we report the mean and standard error of each metric across 10 shots, using a fixed set of hyperparameters for the QUBO model across all PDBs. The number of feasible results per PDB are 3WNE = 10, 5LSO = 10, 3AV9 = 4, 3AVI = 6, 3AVN = 8, 2F58 = 2.
MJ interactions | RMSD (Å) | ||||||
|---|---|---|---|---|---|---|---|
PDB ID | QUBO | CP (1-NN) | CP | QUBO | CP (1-NN) | CP | \({t_\text {QUBO}}\) (s) |
3WNE | \(-1.87\pm 0.00\) | \(-1.87\pm 0.00\) | \(-46.30\pm 0.00\) | \(10.03\pm 0.20\) | \(7.77\pm 0.10\) | \(7.51\pm 0.00\) | \(126\pm 7\) |
5LSO | \(-5.93\pm 0.04\) | \(-5.97\pm 0.00\) | \(-53.56\pm 0.00\) | \(13.31\pm 0.16\) | \(13.20\pm 0.13\) | \(13.60\pm 0.06\) | \(161\pm 11\) |
3AV9 | \(-6.18\pm 0.40\) | \(-6.64\pm 0.00\) | \(-101.09\pm 0.00\) | \(11.36\pm 0.16\) | \(9.92\pm 0.27\) | \(9.56\pm 0.20\) | \(156\pm 16\) |
3AVI | \(-7.45\pm 0.64\) | \(-8.54\pm 0.00\) | \(-114.71\pm 0.00\) | \(9.65\pm 0.52\) | \(9.39\pm 0.04\) | \(8.87\pm 0.00\) | \(191\pm 18\) |
3AVN | \(-6.52\pm 0.21\) | \(-6.88\pm 0.00\) | \(-102.01\pm 0.00\) | \(10.60\pm 0.45\) | \(10.75\pm 0.26\) | \(12.28\pm 0.16\) | \(325\pm 15\) |
2F58 | \(-6.18\pm 0.11\) | \(-13.61\pm 0.00\) | \(-134.19\pm 0.00\) | \(11.52\pm 0.33\) | \(11.39\pm 0.38\) | \(10.83\pm 0.30\) | \(237\pm 23\) |
