[phenixbb] sum(occ)<1?

Ed Pozharski epozh001 at umaryland.edu
Fri Jul 16 09:07:30 PDT 2010

On Thu, 2010-07-15 at 15:42 -0700, Pavel Afonine wrote:
> It is in todo list, but it doesn't have high priority. Frankly, I
> don't 
> know how to do this cleanly - minimization with inequality constraint 
> (as opposed to what we have no - minimization with strict
> constraint). 
> I'm sure there are right methods for this, but it might turn into a
> few 
> months project to accomplish for unclear gain.

Reset occupancies that are outside the [0,1] range at every step of
minimization.  Even better approach is to revert back to the previous
step, fix the offending parameters at 0 or 1 and then minimize using the
rest of the parameters.  You can also try adding a steep penalty to the
target function - this will allow occupancies to be a tiny bit outside
the range during minimization but the offending parameters can be reset
and fixed prior to the very last step.

A cheaper trick would be to create a dummy atom that contributes nothing
to Fc (just set its scattering factor to zero) and then add it to the
group as a "sink".

Others may come up with better ideas but I am not trying to compel
anyone to actually implement anything, since it may turn into a few
months project for me ;)  Careful reading of my post shows no such
requests, and in fact there is an alternative: latest version of refmac
does allow refining occupancy for multiple conformers without
constraining total occupancy to 1.

"I'd jump in myself, if I weren't so good at whistling."
                               Julian, King of Lemurs

More information about the phenixbb mailing list