Application of Enhanced Sampling Monte Carlo Methods for High-Resolution Protein-Protein Docking in Rosetta

Abstract

The high-resolution refinement of docked protein-protein complexes can provide valuable structural and mechanistic insight into protein complex formation complementing experiment. Monte Carlo (MC) based approaches are frequently applied to sample putative interaction geometries of proteins including also possible conformational changes of the binding partners. In order to explore efficiency improvements of the MC sampling, several enhanced sampling techniques, including temperature or Hamiltonian replica exchange and well-tempered ensemble approaches, have been combined with the MC method and were evaluated on 20 protein complexes using unbound partner structures. The well-tempered ensemble method combined with a 2-dimensional temperature and Hamiltonian replica exchange scheme (WTE-H-REMC) was identified as the most efficient search strategy. Comparison with prolonged MC searches indicates that the WTE-H-REMC approach requires approximately 5 times fewer MC steps to identify near native docking geometries compared to conventional MC searches.

Fig 1. Docking refinement conditions.

Each docking starting geometry was generated by an initial random translation of one unbound partner from the geometry in the complex by 15Å and random rotation of 60° (compare green displaced and grey cartoon representations). During the docking search translation and rotation of one partner with respect to the other was restricted relative to the starting geometry by 20Å and 90° (indicated by red circle), respectively.

Fig 2. Workflow represented by combination of Rosetta modules and setup of the four docking protocols.

The modules in orange, representing the enhanced sampling techniques of replica exchange and well-tempered ensemble, are only applied in the combined protocols.

Fig 3. Scatter plot of interaction score I_sc (Rosetta units) vs. L_rmsd (Å) for the four docking refinement protocols and two representative targets, 1EAW (A) and 3SGQ (B).

The protocol is indicated on the left for each row of plots. The snapshots number is color-coded, that means blue and red dots corresponding to decoys sampled at the beginning and the end of the docking searches in each panel, respectively. The three columns of plots indicate the result after different simulation lengths (indicated on top of each column).

Fig 4. Agreement between sampled docking geometries and the corresponding bound complex.

(A, upper panel) Highest fraction of native contacts (f _nat) found in the top 10 decoys (according to L_rmsd) sampled in each protocol (2x10⁶ MC steps). (A, lower panel) Fraction of CAPRI medium and high quality complexes found for each target and each protocol (the protocols MC, REMC, WTE-REMC and WTE-H-REMC are indicated by different colors). (B) same as in (A) but for the docking refinement runs with 10⁷ MC steps.

Fig 5. Evolution of I_sc docking interaction score (A) and best sampled Lrmsd (B) vs. MC step number.

The MC step number is scaled by x1000. For the interaction score I_sc the smallest difference (sampled up to the selected step number) relative to the lowest scoring complex sampled in the entire docking search is plotted. The variance in sampled scores (up to the considered number of MC steps) is indicated by error bars for the MC protocol. It is of similar magnitude for the other protocols (not shown). For (B) the smallest sampled L_rmsd up to the step number indicated in the x-axis is shown.