The following two figures compare a 4.6Å thick stack of sections from
a 4.0-0.8Å
(2Fo - Fc)exp(i
)
GraphEnt (top) and conventional (lower) maps for a 4-
-helical bundle
protein at the final stage of its refinement (R = 0.102).
Both maps are contoured at 1.3
above their respective mean
densities.
As it is obvious, the conventional map suffers from quite appreciable series termination errors and contains a number of peaks (and continuous features) that are not required by the data. If you think that this is understandable (given that the low resolution cutoff was only 4Å), re-consider : the low resolution data that have been excluded from the calculation are only 204 reflections out of a total of 48774 reflections (ie less than 0.4% of the total number of reflections).
