DARC 2.0: Improved Docking and Virtual Screening at Protein Interaction Sites
Fig 4
Screening ligand conformers “on-the-fly” during docking.
(A) Previously, ligand conformations were docked sequentially through separate docking trajectories, and the ligand conformation was ultimately drawn from the best-scoring complex. In DARC 2.0, we instead sample ligand conformers during the docking trajectory. (B) Due to the extra degree of freedom associated with a single docking trajectory, docking converges more slowly when conformers are sampled “on-the-fly”. Here, convergence is evaluated by the score difference relative to a “gold standard” (best achievable score) for each complex; the results shown are averaged over the 25 complexes in our test set (S1 Table). For each point in this plot, the number of particles in PSO optimization and the number of steps in the docking trajectory were set equal to one another. (C) Despite the fact that individual trajectories converge more slowly when conformers are sampled “on-the-fly”, docking is comprised of only a single trajectory. Across our set of 25 complexes, this ultimately makes “on-the-fly” sampling an average of 18-fold faster than sampling conformers sequentially through multiple docking trajectories.