This makes it possible to calculate the resolution of a structure regardless of how it was solved (X-ray, NMR, EM, modeling, ab initio prediction).
[1] Resolution-by-proxy values can be calculated for newly determined protein structures using a freely accessible ResProx web server.
[1] This server accepts protein coordinate data (in PDB format) and generates a resolution estimate (in Angstroms) for that input structure.
In X-ray crystallography, resolution is a measure of the resolvability or precision in the electron density map of a molecule.
When the resolution is greater than 3.5 Å, there is often considerable uncertainty in both the atom locations and even the identity of individual amino residues.
Likewise, in NMR it is possible to generate high quality, precisely determined protein structures using relatively few, well-chosen constraints.
It is also possible to generate very low quality NMR structures from large numbers of carelessly assessed, mistaken or mis-assigned constraints.
[1] After training and appropriate validation on independent tests sets, this SVR model is able to estimate the resolution of solved X-ray structures with a correlation coefficient of 0.92, mean absolute error of 0.28 Å.
Because the performance of the ResProx method is so high and because it only needs coordinate data to generate an estimate of the equivalent X-ray resolution, it is ideally suited to be applied to NMR structures.
As seen in Figure 2, a very small number NMR structures exhibit a resolution equivalent to < 1.0 Å, but these are rare.