Vina [82] (see Note 15). The input for the docking software is
generally unique for that program. However, there are common
steps necessary for most software, which include the preparation of
receptor files, ligand file, choice of scoring function (some programs such as GOLD contain multiple scoring functions), and
receptor flexibility, if allowed. If there is experimental evidence
which residues are involved in ligand binding, it may be beneficial
to set these residues as flexible, else rigid docking is advised.
The output of docking software is usually given as a set of
predicted ligand-binding conformations. To interpret this, conformations predicted by an individual program and scoring function
should be clustered. Once clustered, conformations that were predicted by multiple programs should be considered first (consensus
docking, see [54]).
To identify ligand conformations that can describe the density
in the map, the goodness-of-fit between the ligand and the map can
be calculated. This can be done with both the ligand-bound conformation map and the difference density map (calculated between
the ligand-bound conformation map and an unbound conformation, either from a map or a model) (see Note 16). Conformations
with a high correlation for both difference and density maps should
also be visually inspected for factors such as electrostatic complementarity between protein and ligand, the fit of the ligand into the
density and ligand-residue interactions.
5.8 Jwalk and MNXL
Jwalk calculates distance between cross-linked residues that can be
used to validate candidate models. Two distance metrics are used,
Euclidean distance and the Solvent-accessible surface distance
(SASD). The SASD has the advantage that it computes the distance
across the protein surface, as opposed to the Euclidean distance
(which does not take into account the fact that cross-linkers cannot
travel through the protein).
Jwalk is available as a web server (see Note 17), the web server
takes a PDB file as input. Two options are available for calculating
SASD: (a) calculation of all SASD between two amino acid types
(either Lysine, Cystine, Glutamate, or Aspartate). (b) calculation of
SASD from an uploaded list of residues (see Note 18). Additionally,
the maximum SASD length and grid spacing (A ˚ ) can be specified.
The output gives the SASD and Euclidean distance for each calculated link and the cross-links are available for download in PDB
format.
The Jwalk output can then be used along with experimental
cross-link data to validate multiple models using the MNXL or
cMNXL scoring functions. The MNXL web server (see Note 19)
takes an input of: (a) experimental cross-link data. (b) precalculated
Jwalk output, or PDB models for validation. The output contains
the MNXL score, along with the number of matched, violating, and
nonaccessible cross-links.
216
Tristan Cragnolini et al.
generally unique for that program. However, there are common
steps necessary for most software, which include the preparation of
receptor files, ligand file, choice of scoring function (some programs such as GOLD contain multiple scoring functions), and
receptor flexibility, if allowed. If there is experimental evidence
which residues are involved in ligand binding, it may be beneficial
to set these residues as flexible, else rigid docking is advised.
The output of docking software is usually given as a set of
predicted ligand-binding conformations. To interpret this, conformations predicted by an individual program and scoring function
should be clustered. Once clustered, conformations that were predicted by multiple programs should be considered first (consensus
docking, see [54]).
To identify ligand conformations that can describe the density
in the map, the goodness-of-fit between the ligand and the map can
be calculated. This can be done with both the ligand-bound conformation map and the difference density map (calculated between
the ligand-bound conformation map and an unbound conformation, either from a map or a model) (see Note 16). Conformations
with a high correlation for both difference and density maps should
also be visually inspected for factors such as electrostatic complementarity between protein and ligand, the fit of the ligand into the
density and ligand-residue interactions.
5.8 Jwalk and MNXL
Jwalk calculates distance between cross-linked residues that can be
used to validate candidate models. Two distance metrics are used,
Euclidean distance and the Solvent-accessible surface distance
(SASD). The SASD has the advantage that it computes the distance
across the protein surface, as opposed to the Euclidean distance
(which does not take into account the fact that cross-linkers cannot
travel through the protein).
Jwalk is available as a web server (see Note 17), the web server
takes a PDB file as input. Two options are available for calculating
SASD: (a) calculation of all SASD between two amino acid types
(either Lysine, Cystine, Glutamate, or Aspartate). (b) calculation of
SASD from an uploaded list of residues (see Note 18). Additionally,
the maximum SASD length and grid spacing (A ˚ ) can be specified.
The output gives the SASD and Euclidean distance for each calculated link and the cross-links are available for download in PDB
format.
The Jwalk output can then be used along with experimental
cross-link data to validate multiple models using the MNXL or
cMNXL scoring functions. The MNXL web server (see Note 19)
takes an input of: (a) experimental cross-link data. (b) precalculated
Jwalk output, or PDB models for validation. The output contains
the MNXL score, along with the number of matched, violating, and
nonaccessible cross-links.
216
Tristan Cragnolini et al.