Figure 1.
The low and high resolution stages are shown in shades of green and blue respectively. The Move Set box illustrates five different trial perturbations which are chosen randomly with indicated frequencies. *Rigid body positions are minimized corresponding to the rigid body perturbation move selected. If all the CDRs are selected (see Move Set), they are minimized. If CDRs H3 or H2 are selected for perturbation, they are not subjected to additional minimization since they are already minimized.
Figure 2.
Docking perturbation plots for blocking antibody 11k2 complexed with human monocyte chemoattractant protein (MCP)-1 (2BDN [33]).
(A) Standard rigid-body docking using RosettaDock starting with the antibody crystal structure. The red point represents the native crystal structure. (B) Standard rigid-body docking using RosettaDock with the lowest-energy RosettaAntibody model. (C) Docking with VL-VH optimization using the lowest-energy RosettaAntibody model. (D) Docking with VL-VH optimization with CDR minimization and CDR H3 perturbation using the lowest-energy RosettaAntibody model. (E) Docking with SnugDock (VL-VH optimization with CDR minimization and CDR H3+H2 perturbations) using the lowest-energy RosettaAntibody model. (F) Rigid-body docking using EnsembleDock with the ten lowest-energy RosettaAntibody models. (G) Docking using a combined protocol incorporating EnsembleDock and SnugDock with the ten lowest-energy RosettaAntibody models.
Figure 3.
Summary of docking performance.
The bar plots show the number of correctly docked targets out of fifteen targets for different docking algorithms. (A) Docking performance considering the lowest-energy decoy. (B) Docking performance considering the most native-like prediction in the ten lowest-energy decoys. (C) Docking performance based on the presence of docking energy funnels. Crystal indicates standard rigid-body docking using crystal structures. RosettaAntibody indicates standard rigid-body docking using RosettaDock starting with the lowest-energy RosettaAntibody homology model. VL-VH indicates docking with VL-VH optimization. VL-VH+CDR H3 indicates docking with VL-VH optimization with CDR minimization and CDR H3 perturbation. SnugDock indicates docking using SnugDock. Rigid Body Ensemble indicates rigid-body docking using EnsembleDock with the ten lowest-energy RosettaAntibody homology models. Snug+Ensemble indicates docking using the EnsembleDock-plus-SnugDock combined protocol with the ten lowest-energy RosettaAntibody homology models.
Table 1.
Accuracy of lowest-energy docking decoy for different docking protocols.
Figure 4.
SnugDock conformational diversity.
(A) The diversity in conformation generated by SnugDock during docking of anti-HEL Fab fragment (1BQL) to bobwhite quail lysozyme (1DKJ). (B) View facing the paratope. Crystal structure, red; heavy and the light chains, blue and yellow, respectively; light and heavy chain CDRs, orange and cyan, respectively; SnugDock sampled CDR H3, grey; EnsembleDock-plus-SnugDock sampled CDR H3, light chain CDRs and light chain framework, green, light orange and yellow-green, respectively. Structures are all superposed onto the heavy chain framework residues of the crystal structure. (C) Mean rmsd from the starting structure of the ten lowest-energy docking decoys for fifteen targets. For light and heavy chain CDRs, the corresponding framework chain is superposed and the rmsd is queried over the respective CDR residues. VL-VH denotes the rigid-body rmsd divergence of the heavy chain framework when the light chain framework is superposed. The paratope comprises all CDRs, and the rmsd was computed by superposing the paratope and querying over the same residues. The colors of the bar correspond to the colors of the different antibody segments in (A) and (B). The error bars denote one standard deviation.
Figure 5.
Structural details of the monoclonal antibody Fab D44.1 complexed with lysozyme (1MLC [38]).
(A) The interface region of the lowest-energy RosettaAntibody homology model for target 1MLB complexed with the crystal structure of lysozyme (1LZA). (B) The interface region of the most native-like prediction in the ten lowest-energy docking predictions on docking with standard rigid-body RosettaDock. (C) The interface region of the most native-like prediction in the ten lowest-energy docking predictions on docking with SnugDock. (D) Superposition of the structures shown in (B) and (C) viewed facing the binding region from the antigen's side. Conformations of the antigen in the crystal structure, green; predicted by standard rigid-body RosettaDock, red; and that predicted by SnugDock, grey; heavy and light chains, shades of blue and yellow respectively. Sticks indicate the labeled residues that have relieved the steric clash present in the starting structure due to the flexibility allowed by SnugDock. Transparent spheres indicate the interface region of the predicted conformation of the antigen. The light and heavy chain frameworks of the predicted complexes are superposed on the corresponding residues of the antibody in the crystal structure.
Figure 6.
Predicted models of the complex of west Nile virus envelope protein DIII with neutralizing E16 antibody Fab (1ZTX [39]).
(A) Lowest-energy prediction (medium accuracy) generated by EnsembleDock-plus-SnugDock simulations ranked by all-atom score of the entire complex. (B) Lowest-energy prediction (acceptable accuracy) generated by EnsembleDock-plus-SnugDock simulations ranked by the intermolecular components of the all-atom score. The light (deep blue) and heavy (yellow) chain framework of the docked antibody is superposed on the corresponding residues of the crystal complex. Predicted orientation of the antigen, green; light and heavy chain CDRs, orange and cyan respectively; CDR H3 loop and antigen in the crystal structure, red; residues at the interface, transparent spheres.
Table 2.
Accuracy of lowest-energy docking decoy, using WAM homology models.
Table 3.
Accuracy of global docking decoys.