9. Validate and estimate model quality
Evaluate and estimate the quality of the structure. You can use the same
tools as for experimental structures. Ramachandran plot can be
investigated and analyzed with several tools. PROCHECK (Laskowski et
al., 1993), WhatCheck (Hooft et al., 1996), MolProbity (Williams et al.,
2018), Verify3D (Lüthy et al., 1992) and many other tools are available.
Several of them can be run via SAVES (https://saves.mbi.ucla.edu/) to
control stereochemical quality, bond angles, bond lengths, etc. At the
Critical Assessment of protein Structure Prediction (CASP) challenges
(https://www.predictioncenter.org/) the community has developed various
tools and measures e.g. to compare structures with experimental
structures.
One way to estimate usefulness of the model is to use it to interpret
experimental data. If successful, the model can be useful also for other
applications. Describe any details of such applications.
Example. The quality of the produced model was evaluated with
PROCHECK (Laskowski et al., 1993) and MolProbity (Williams et al.,
2018). Ramachandran plot (Ramachandran et al., 1963) indicated that 91%
of the residues are on the most favored area and only two in disallowed
region. Bond lengths and angles are well within normal ranges (on
average 0.011Å and 2.1 degrees, respectively), the same with the torsion
angles. According to MolProbity there are no clashes and there are not
issues with geometry. The model was tested also by the Verify3D
technique (Lüthy et al., 1992) and found to have the score of a typical
globular protein.
As as functional test we evaluated the method in describing a known
interaction based on a amino acid substitution. D304L in the lower lobe,
was explained based on the model as having a structural effect
originating from the disruption of a salt bridge with R244 in the upper
lobe. The prediction is correct based on experimental
data.