Challenges in Modelling Metalloenzymes
505
2 Checking the Structure Quality
Computational studies on metalloenzymes depend heavily on available experimental
macromolecular structures, and in most cases, one starts his/her project by downloading a suitable structure from the Protein Data Bank (PDB). Irrespective of if the
structure available at PDB is for the actual enzyme of interests or it will be used as a
template in homology modelling, it is highly desirable that it is complete and accurate, especially in the vicinity of the active site hosting the metal(s) ion(s). The overall
quality of the structure can be inferred from various metrics reported at the PDB web
page, such as resolution, R-value free, R-value work, clashscore, Ramachandran outliers and other, yet these global parameters take into account thousands of atoms of
the structure, of which only a few are metal ions; hence, their informative value as
to the quality of the structure in the immediate vicinity of the metal is rather limited. This is compounded by the fact that force fields commonly used for structure
refinement are more accurate for standard amino acids and organic molecules than
for inorganic complexes. Thus, it is advisable to invest some time and check the
quality of the metal-binding site before undertaking some more demanding modelling tasks. CheckMyMetal (CMM, https://csgid.org/metal_sites) is an easy to use
metal-binding site validation server that can be readily used for this purpose [8, 9].
For a given macromolecular structure, specified either by its PDB four-letter code
or uploaded in the PDB format, CMM computes seven quality parameters derived
solely from the geometry of the first coordination shell and two metrics related to the
experimental model. Computed values are compared to reference ranges (Acceptable, Borderline and Outlier) that were derived based on benchmark data of 7350
metal-binding sites from 2304 high-resolution crystal structures.
As an illustrative example of using CMM may serve an analysis of a metal-binding
site in the structure of α-galactosidase (PDB: 3LRK), even though in this case the
metal is most likely not involved in the enzymatic reaction [8, 10]. 3LRK has been
solved with a good resolution (1.95 Å), and it also shows rather good values for
other metrics reported by the PDB server. In the original structure, one Na
+ cation
was modelled to be coordinated by His98, His102, one nitrogen and four oxygen
atoms from a bis-tris buffer molecule. Analysis of this site by CMM shows that three
parameters (Ligands, Valence and Geometry) fall into the Outlier range, whereas
other two (B factor and nVECSUM) are classified as Borderline. Further inspection
of electron density difference maps revealed that sodium cation cannot account for
the observed electron density peak, and the site was re-refined with Cu
+ instead of
Na
+ and the bis-tris molecule replaced by two glycerol and two water molecules, all
four with partial occupancy of 0.5 (Fig. 1).
As can be noted in the figure, the curated model successfully accounts for the
observed electron density and in the CMM analysis performed for this model all
quality metrics fall either into Acceptable or Borderline ranges. Detailed description
of the CMM protocol, its limitations (e.g. the valence model does not cover metal–metal, delocalized and π metal–ligand bonds) and further examples of use can be
found in the original CMM papers [8, 9].
505
2 Checking the Structure Quality
Computational studies on metalloenzymes depend heavily on available experimental
macromolecular structures, and in most cases, one starts his/her project by downloading a suitable structure from the Protein Data Bank (PDB). Irrespective of if the
structure available at PDB is for the actual enzyme of interests or it will be used as a
template in homology modelling, it is highly desirable that it is complete and accurate, especially in the vicinity of the active site hosting the metal(s) ion(s). The overall
quality of the structure can be inferred from various metrics reported at the PDB web
page, such as resolution, R-value free, R-value work, clashscore, Ramachandran outliers and other, yet these global parameters take into account thousands of atoms of
the structure, of which only a few are metal ions; hence, their informative value as
to the quality of the structure in the immediate vicinity of the metal is rather limited. This is compounded by the fact that force fields commonly used for structure
refinement are more accurate for standard amino acids and organic molecules than
for inorganic complexes. Thus, it is advisable to invest some time and check the
quality of the metal-binding site before undertaking some more demanding modelling tasks. CheckMyMetal (CMM, https://csgid.org/metal_sites) is an easy to use
metal-binding site validation server that can be readily used for this purpose [8, 9].
For a given macromolecular structure, specified either by its PDB four-letter code
or uploaded in the PDB format, CMM computes seven quality parameters derived
solely from the geometry of the first coordination shell and two metrics related to the
experimental model. Computed values are compared to reference ranges (Acceptable, Borderline and Outlier) that were derived based on benchmark data of 7350
metal-binding sites from 2304 high-resolution crystal structures.
As an illustrative example of using CMM may serve an analysis of a metal-binding
site in the structure of α-galactosidase (PDB: 3LRK), even though in this case the
metal is most likely not involved in the enzymatic reaction [8, 10]. 3LRK has been
solved with a good resolution (1.95 Å), and it also shows rather good values for
other metrics reported by the PDB server. In the original structure, one Na
+ cation
was modelled to be coordinated by His98, His102, one nitrogen and four oxygen
atoms from a bis-tris buffer molecule. Analysis of this site by CMM shows that three
parameters (Ligands, Valence and Geometry) fall into the Outlier range, whereas
other two (B factor and nVECSUM) are classified as Borderline. Further inspection
of electron density difference maps revealed that sodium cation cannot account for
the observed electron density peak, and the site was re-refined with Cu
+ instead of
Na
+ and the bis-tris molecule replaced by two glycerol and two water molecules, all
four with partial occupancy of 0.5 (Fig. 1).
As can be noted in the figure, the curated model successfully accounts for the
observed electron density and in the CMM analysis performed for this model all
quality metrics fall either into Acceptable or Borderline ranges. Detailed description
of the CMM protocol, its limitations (e.g. the valence model does not cover metal–metal, delocalized and π metal–ligand bonds) and further examples of use can be
found in the original CMM papers [8, 9].