Example: Use of a Solvent Mask for Bulk Solvent Correction

Since the bulk solvent density is not usually known, the solvent structure factors must be scaled to the FCALC terms. Also, the grid being transformed has a sharp edge between the solvent and solvent-excluded regions, so Fourier ripples are generated that adversely affect the high-resolution terms. A large temperature factor can be applied to smooth these edge effects. Thus, a DO statement, ${\rm FPART} = {\rm FPART} \; k_{solvent} \; exp(-B_{solvent}s^2/4)$, is used to correct the calculated solvent structure factors in FPART. $k_{solvent}$ and $B_{solvent}$ are determined empirically by iteratively searching for the value of one parameter that minimizes the $R$ value in the lowest-resolution shell without significantly increasing the high-resolution $R$ values, keeping the other parameter fixed. Typically, initial values of $k$=0.40 e$^-$ Å$^{-2}$ and $B$=200 Å$^2$ are chosen. The X-PLOR shell language is used to loop over values of $k$ in a coarse search; the optimum value of $k$ is subsequently used in a coarse search to optimize $B$. This procedure is subsequently repeated over a finer set of values for $k$ and $B$. All search procedures have to be done manually.

The following example shows how to compute the solvent mask:

solmask.inp

The next example shows how to test various scale and B-factors. Note that this protocol has to be run several times with different trial scale factors and B-factors. One should look at low $R$ values for the low-resolution bins while maintaining low $R$ values for the high-resolution bins. This is a somewhat subjective procedure, which is the reason why this protocol is not automated.

solscale.inp

In subsequent protocols, one should include the following lines before computing $R$ values, crystallographic targets, or electron density maps:

reflection @amy_s end            {*Reflection file with solvent mask FPART.*}
resolution 40. 2.
do ( FPART= 0.4 * exp(-200*(s()^2)/4.)*FPART )