Table 2.

A summary of studies integrating HDX-MS and docking to predict Ab-Ag native poses and their performance.

Studies	Docking algorithms	Docked Ag	Docked Ab	Sampling and applying HDX restraints	Model selection	Evaluation
(27)	ZDOCK (in combination with ZRANK and RDOCK)	crystal structure	crystal structure (Fab) or homology modeled antibody (Fv) using RosettaAntibody protocol	-54000 poses were sampled in ZDOCK -All residues in non-CDR regions (Ab) and in non-epitope peptides (Ag) were blocked during ZDOCK, allowing the pairwise shape complementarity (PSC) scoring function of ZDOCK to penalize docked poses when the blocked residues are in the interface -The initial-stage docking poses were scored and ranked using ZRANK -The top 50 rigid-body docking poses were refined using RDOCK	-The top 10 poses from RDOCK refinement process were selected for evaluation -The RMSD value of the interface Cα atoms (iRMSD) was calculated by superimposing a docked pose onto the relax co-crystallized antigen-antibody complex. Docking poses with an iRMSD less than or equal to 2.5 Å are considered near-native structures or “hits” with the interface defined as all residues with at least one atom within 10 Å of the binding partner	-Compared to stand-alone docking, the HDX-MS-derived restraints significantly improved the docking results for one of the three testing Ab-Ag complexes: the number of “hit” poses among top 10 poses generated increased from four to seven, with the iRMSD of the highest-ranking pose being 1.4 Å to the complex crystal structure -Incorporation of HDX-MS data produced more tightly clustered docking poses for all three complexes and did not interfere with the result when the stand-alone docking already did well by itself
(32)	MOE	crystal structure	models of HDX-predicted antibody peptides were generated with MOE	-HDX-predicted epitope peptides were set as the docking sites	-Five (MOE) to ten (PatchDock and ZDOCK) molecular dynamics-minimized docking poses were selected for evaluation -Optimal poses are those with the highest numbers of energetically favorable contacts (“hit”) between the paratope peptides and the antigen, where a “hit” is regarded when a residue from a predicted epitope peptide located within 4.5 Å of a residue from the antibody peptide	-For all three software packages, computational docking with HDX-MS data produced more “hit” residues than docking without HDX-MS data. In other words, more ‘hit’ residues were detected for docking at the HDX-specified site compared to randomly selected sites -The crystal structure of the Ab-Ag complex is not available. Thus, it cannot be determined how much the iRMSD to the native structure improved with the incorporation of HDX-MS
	PatchDock			-HDX-predicted epitope peptides were set either as the docking sites or as volume-constraint pharmacophores
	ZDOCK			-HDX non-epitope residues of the antigen were blocked as a scoring penalty
(11)	MOE	crystal structure	Homology modeled antibody (Fab) using Bioluminate protocol v1.9 and MAESTRO v10.2	-100,000 starting poses were sampled using MOE -CDR restraints were applied by using an energy penalty to require that all poses contain a minimum number of residue contacts between HDX-predicted paratope peptides and these regions -The poses were further refined and scored using a full-atom potential (AMBER)	-The top 200 poses were evaluated for surface complementarity based upon AMBER complementarity score (24) and visual inspection of surfaces as implemented in the protein_contact_surfaces script implemented in MOE	-The best docking poses were proposed to be the Ab-Ag interaction model. The HDX-predicted peptides in this model were at the interface and were corroborated by the SASA analysis -No blind docking was done for parallel comparison
(14)	Rosetta	crystal structure	crystal structure (Fab)	-Restrict docking in Rosetta to HDX-predicted epitope of the antigen and the CDRs region of antibody -The docking poses were filtered by overall energy, binding energy, and satisfaction to HDX constraints -The best 500 models by binding energy underwent the protocol again	-An ensemble of 25 best-scoring models (by binding energy) that fulfilled HDX constraints were selected	-The best docking poses were proposed to be the Ab-Ag interaction model. Functional assays were performed, and the results endorsed the binding modes of the docked complexes -No blind docking was done for parallel comparison
(36)	PatchDock	crystal structure	Homology modeled Ab (Fv) using ABodyBuilder Fv prediction	-HDX-predicted epitopes were set as docking sites by adding a scoring parameter to PatchDock -The clustering RMSD was set at 4 Å	-The top 100 poses were evaluated for CDR inclusion at the interface and agreement to the alanine scan data. Among these, the top two poses were selected	- HDX profile simulation was performed using the ‘calc-HDX’ function of the HDXer tool for the top two docked structures -Comparison between the simulated ΔHDX versus the experimental ΔHDX further validated these poses, displaying a RMSD of deuterium exchange of 0.981 Å and 0.684 Å -No blind docking was done for parallel comparison