***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***


Nicholas M Glykos <[EMAIL PROTECTED]> wrote:

> To give a semi-philosophical argument, the best model is the one that 
> maximises the product (likelihood * prior). A model for which all surface 
> exposed lysines, arginines, etc, have all atoms after Cbeta missing should 
> be given an extremely low prior probability (based purely on chemical- 
> biological grounds :-)

I strongly agree with this viewpoint.
We should not be discussing _whether_ to include these atoms in the model,
but how _best_ to include these atoms in the model.

: 
Marc Delarue <[EMAIL PROTECTED]> wrote:

> One possibility is indeed to put all possible rotamers coordinates, only I
> would not recommand 1/Nrot occupancy, but, instead, Boltzmann-like weights
> with an Energy taking into account both the Corr. coeff. of that
> particular rotamer in the electron density (E-1-CC) and a penalty given by
> some oracle called Ms_Bump.

I like this idea, but you would have to watch out for a proliferating number
of parameters to implement it.   If you want separate occupancy, weights, and
B factors for each rotamer, that is a suspiciously large number of parameters
in the case of Arg or Lys.


On Wednesday 10 January 2007 09:01, Marc SCHILTZ wrote:
> 
> So the criterion of atoms that should be modelled because they "are in 
> the crystal" has now been narrowed down to atoms that should be modelled 
> because they are "covalently connected to the rest of the protein".
> 
> If this concept is taken further, one could ask why we are not 
> explicitly refining hydrogen atoms ? 

We are.  
At least, those who correctly use available options in refmac or shelx are.
The best results (R, Rfree, geometry) are obtained by explicit inclusion
of hydrogens via the riding-hydrogen model. This is also a basis for the
Molprobity validation tools.

> But we are not refining hydrogen atoms because the experimental data 
> does not warrant this ! (unless we got very high-resolution data). 

Not true.  In fact the hydrogen scattering contribution falls off with
resolution just like the normal scattering from any other atom type.

Reply via email to