Coot provides several graphs that are useful for model validation (on a residue by residue basis): residue density fit, geometry distortion, temperature factor variance, peptide distortion and rotamer analysis.
The density fit graph shows the density fit for residues. The score is the average electron density level at the atom centres of the atoms in the residue. The height of the blocks is inversely proportional to the density average.
The residue density fit is by default scaled to a map that is calculated on the absolute scale. Sometimes you might be using a map with density levels considerably different to this, which makes the residue density fit graph less useful. To correct for this you can use the scripting function:
(set-residue-density-fit-scale-factor factor)
where factor would be 1/(4\sigma_map) (as a rule of thumb).
(residue-density-fit-scale-factor) returns the current scale
factor (default 1.0).
There is also a GUI to this:
Extensions -> Refine... -> Set Density Fit Graph Weight...
Residue rotamers are scored according to the prior likelihood. Note that when CD1 and CD2 of a PHE residue are exchanged (simply a nomenclature error) this can lead to large red blocks in the graph (apparently due to very unlikely rotamers). There are several other residues that can have nomenclature errors like this. To fix these problems use
(fix-nomenclature-errors imol)
This idea is from Eleanor Dodson, who liked to use the standard deviation of a residue's temperature factors to highlight regions of questionable structure.
Note that Hydrogens are ignored in this analysis.
Some variability of the \omega is to be expected in the peptide bond. But not too much. Anything more than 13 degrees is suspicicous. Unexpected peptide bonds show up red by default. If cis peptides are to be expected, and should not marked as bad, then you can tell this to Coot using:
Edit -> Preferences -> Geometry -> Cis-Peptides -> No