[buster-discuss] Bulk solvent model

[Dale Tronrud]

Hi,

   I'll jump in here since I know a little about the local scaling
method of bulk solvent modeling.  Your understanding of the meaning
of Ksol and Bsol is correct.  With T4 lysozyme and 2M PO4 mother
liquor I've seen Ksol's around 0.9 while with Thermolysin with a
low ionic strength mother liquor I've seen it around 0.7.  Without
high resolution data to help the calculation, however, the
uncertainties of these parameters can be quite large.  With 4 A
data I'm not surprised at all that you see a ksol this large.

   On the plus side, if your data has insufficient resolution
breadth to allow a solid estimation of Ksol it is probably not
broad enough to be hurt by the inaccuracies of the bulk solvent
model.  The idea of Babinet scaling is that the low resolution
and high resolution data need different scaling, but if you only
have low resolution data there isn't as much of a problem.

   I don't expect much difference between you sticking with 1.0
or forcing Buster to use a value of 0.9 or 0.8.  Removing the
bulk solvent model by setting Ksol to 0.0 will make things worst.

   A model that is incomplete by missing side chains will cause
an underestimate of the average e/A^3 in the protein region and
a corresponding underestimate of the density in the solvent region,
but the Babinet scaling should still get the ratio right.  When
the model is missing domains Babinet scaling will tend to lower
the difference density in those regions because bulk solvent is
modeled there.  If you think this is a serious problem you need
to use a mask method of solvent modeling and manually exclude
these regions from the solvent mask.  I don't think very many
people actually do this.

   The only correlation I have seen between Ksol and resolution
is that low resolution refinements tend to have Ksols closer
to 1.0, but that is simply due to the lack of data causes an
inability to accurately determine Ksol.

Dale Tronrud

On 02/16/11 07:38, Kenneth Verstraete wrote:
> Dear all,
> 
> I have been working on the low resolution refinement (4.3 Angstrom) of
> a complex with autoBuster. I have some questions regarding the
> interpretation of the determined bulk solvent parameters (Ksol and
> Bsol). According to the literature, two solvent models are currently
> used by different refinement packages: the exponential scaling model
> (application of Babinet's principle) and the mask model. According to
> Blanc et al., 2004 (PMID: 15572774) it seems that Buster uses
> Babinet's principle. In this model Ksol is interpreted as the ratio of
> the solvent electron density to the protein electron density, with
> average values of 0.75 to 0.95. Bsol limits the contribution of the
> solvent to > 5? Angstrom and has values of 150 to 350 A^2.
> 
> The values I obtained are: Ksol = 1.032 and Bsol= 155.  Thus it seems
> that the average density of the solvent is higher then that of the
> protein. Should I be worried about this?
> Because of the low resolution most residues are modelled as Ala, and
> even some domains are missing (about 20% of the total protein mass in
> the asu). I wonder if it is reasonable to state that Ksol is unusually
> high because of the unmodelled protein regions? Or is there another
> reason for the high Ksol (e.g. chemical composition of the
> crystallization solution)?
> I also wonder if there is a correlation between Ksol and the
> resolution of the data?
> 
> Thanks for your input,
> 
> best regards,
> 
> Kenneth
> 
>