[phenixbb] Tracing and refining b-strands at 3.3A resolution

[Ralf W. Grosse-Kunstleve]

Hi Phil,

> My question with this approach is that is there any potential to have a 
> negative effect on refining the structure, given that one is introducing 
> a number of bonds with relatively high average deviations from the mean 
> ?  The rest of the bonded (covalent) geometry has an rms in the range 
> 0.004 - 0.015 (-ish) but this is one to two orders of magnitude tighter 
> than a hydrogen bond.  Is the target function sophisticated enough to 
> handle this or is there a problem with treating hbonds and covalent 
> bonds as peers ?

I think the L-BFGS algorithm we're using is smart enough to handle this
situation, but I haven't investigated this question systematically.
Note that e.g. most dihedral restraints are also much weaker than
the bond restraints.

Ralf