Matt Warkentin wrote: > hi folks > > i've almost got my B-factors questions sorted out. one thing i still > don't understand is this: > > in truncate's wilson plot, what is plotted on the vertical axis is not > ln I_mean, but ln (I_mean/ff_mean) > > and ff_mean (or Mn_ff, as truncate calls it) is the "average expected > value of ff". > > my question is how is this value calculated? its resolution > dependence obviously affects the B-factor, so i would like to > understand at least what goes into it. at this point in the analysis, > all truncate knows is the cell/spacegroup, and the approximate number > of AAs right? is this just an expectation of how the amplitudes ought > to fall off vs spacial frequency, without knowing anything about the > detailed structure?
The calculation is exactly that. One assumes that each atom scatters X-rays based on its elemental type/charge, that all B factors are equal, and that the atoms are randomly distributed in the unit cell. Since the atoms are not randomly arranged in protein crystals (they having big solvent channels) the curve deviates from linearity at low resolution. It does work pretty good from about 4.5A resolution and higher. If your molecule has some floppy parts and some rigid parts the curve will be a little concave instead of straight. A common mistake is to compare the Wilson B to the average of the B's in your refined PDB file. Because the Wilson B is based on the high resolution end of the plot, atoms with low B factors are given higher weight in the calculation. Whenever your molecule has a spread of B factors, the Wilson B will be lower than the average of the B factors. If you want to make a proper comparison as a check, you should calculate the Wilson B of your Fcalc's and compare that to the Wilson B of your Fobs's. With any properly refined model these two numbers will be nearly identical. In fact, it's a pretty boring check, since I don't know of any program that would fail. > > thanks! Dale
