Structure Assessment
QMEAN is a common method to estimate the quality of a structure. Based on statistical potentials derived from high-quality
experimental data, it provides an estimate of the quality of the
structure based on the similarity of its geometry to other structures.
[24] Other similar potentials, such as DOPE [66] and ProQ3D
[67] are in common use [68].
Model Validation by
Additional Experiments
To resolve the structures of macromolecular assemblies, multiple
experimental methods may be used, and often provide complementary information [69]. By integrating these different sources, it is
possible to obtain higher accuracy models. However, some of these
experiments can also be used for the purpose of validation, rather
than in the actual model generation itself. In TEMPy, one can
utilize information from cross-linking mass spectrometry for the
validation of the models generated by density fitting.
Jwalk and MNXL: Model
validation Using
Cross-Linking Mass
Spectrometry
Cross-linking mass spectrometry recover information regarding the
distance between residues in a complex. This can be used to filter
out incorrect conformers, or to validate a proposed model. A
compound with two reactivate sites bind with the protein, forming
covalent bonds at two distinct sites. MS analysis then recovers the
position in the sequence of the amino acids involved.
Using a proposed atomic structure, the Jwalk method computes potential paths along the surface of the molecule between
those amino acids using a Breadth-First Search (BFS) algorithm. By
comparing the path length to the cross-linker’s end-to-end distance, one can determine if the model fits with the available crosslinking information [70].
The Matched and Nonaccessible Crosslink (MNXL) score [71]
also includes information regarding the surface accessibility of the
lysines putatively involved in a cross-link: matching pairs where
either end is buried below the protein’s surface are penalized
since they are not solvent-accessible and should not be able to
form cross-links. For each cross-link, the Solvent-Accessible Surface
Distance (SASD), computed as the length of the Jwalk path, is used
to compute a score: a penalty of À0.1 is applied if the path length is
above a certain cutoff (33 A ˚ ), or if the end residues are buried. The
cross-link is then scored using a normal distribution calculated from
all SASDs <33 A ˚ from a cross-linking database.
This score was then extended to maximize its performance
when modeling protein complexes, by treating intra- and interunit cross-links differently; the new score, cMNXL, is described in
[72] (Fig. 10). Both Jwalk and MNXL are available as web
servers [71].
CryoEM Density Fitting and Validation
209
QMEAN is a common method to estimate the quality of a structure. Based on statistical potentials derived from high-quality
experimental data, it provides an estimate of the quality of the
structure based on the similarity of its geometry to other structures.
[24] Other similar potentials, such as DOPE [66] and ProQ3D
[67] are in common use [68].
Model Validation by
Additional Experiments
To resolve the structures of macromolecular assemblies, multiple
experimental methods may be used, and often provide complementary information [69]. By integrating these different sources, it is
possible to obtain higher accuracy models. However, some of these
experiments can also be used for the purpose of validation, rather
than in the actual model generation itself. In TEMPy, one can
utilize information from cross-linking mass spectrometry for the
validation of the models generated by density fitting.
Jwalk and MNXL: Model
validation Using
Cross-Linking Mass
Spectrometry
Cross-linking mass spectrometry recover information regarding the
distance between residues in a complex. This can be used to filter
out incorrect conformers, or to validate a proposed model. A
compound with two reactivate sites bind with the protein, forming
covalent bonds at two distinct sites. MS analysis then recovers the
position in the sequence of the amino acids involved.
Using a proposed atomic structure, the Jwalk method computes potential paths along the surface of the molecule between
those amino acids using a Breadth-First Search (BFS) algorithm. By
comparing the path length to the cross-linker’s end-to-end distance, one can determine if the model fits with the available crosslinking information [70].
The Matched and Nonaccessible Crosslink (MNXL) score [71]
also includes information regarding the surface accessibility of the
lysines putatively involved in a cross-link: matching pairs where
either end is buried below the protein’s surface are penalized
since they are not solvent-accessible and should not be able to
form cross-links. For each cross-link, the Solvent-Accessible Surface
Distance (SASD), computed as the length of the Jwalk path, is used
to compute a score: a penalty of À0.1 is applied if the path length is
above a certain cutoff (33 A ˚ ), or if the end residues are buried. The
cross-link is then scored using a normal distribution calculated from
all SASDs <33 A ˚ from a cross-linking database.
This score was then extended to maximize its performance
when modeling protein complexes, by treating intra- and interunit cross-links differently; the new score, cMNXL, is described in
[72] (Fig. 10). Both Jwalk and MNXL are available as web
servers [71].
CryoEM Density Fitting and Validation
209