[phenixbb] ADP restraints - distance power, average power, etc

Ralf W. Grosse-Kunstleve rwgk at yahoo.com
Sun Dec 19 14:55:12 PST 2010

Hi Joe,
Below are some notes regarding our ADP restraints. At your resolution the 
defaults may be too tight.
The choice of scattering table shouldn't matter, except for small runtime 
differences. (You can find some notes here:
section 4.)

We use "local sphere restraints".

  adp_restraints {
    iso {
      sphere_radius = 5.0
      distance_power = 1.69
      average_power = 1.03

The basic idea is to restrain each adp to the average of all its
neighbors within a sphere of a given radius (sphere_radius = 5). The
contribution to the refinement target function is:

                                   (u_i - u_j)**2
  1 / (r_ij ** distance_power) * ----------------------------------
                                  ((u_i + u_j)/2) ** average_power

These terms are computed over a double sum: loop over each atom, loop
over all neighbors of the atom. I'm not sure anymore how exactly we
arrived at distance_power = 1.69 and average_power = 1.03.
You can try different values for distance_power to change the
tightness of the restraints as a function of the distance of
a pair of atoms. The average_power links the tightness to the
absolute value of the adp; average_power=0 turns this feature off.

There are some remarks about this in the "ADP refinement" section in
this old newsletter article:


The formulas are a bit different (above is current), but the ideas
still apply.

>From: Joseph Noel <noel at salk.edu>
>To: phenixbb at phenix-online.org
>Sent: Sun, December 19, 2010 12:58:43 PM
>Subject: [phenixbb] ADP restraints - distance power, average power, etc
>Hi All,
>I am refining a very well ordered structure at 1.45 A and find that after doing 
>just about everything including optimization of weights, etc that there are 
>still areas of positive density residing over S atoms of Met, backbone atoms, 
>etc. I have added hydrogens and used them during refinement as well. The maps 
>are of a high quality and Free R factors are quite good, ~ 17%. Is there 
>anything that can be played with more such as values used in the ADP restraints 
>window to try and achieve a difference map with far fewer areas of significant 
>positive density (all greater then or equal to 4 sigma)?
>I am not sure exactly what effect the values of distance power, average power, 
>etc will have on refinement. The other option I was thinking of was the 
>scattering table options. I have other structures of this protein that extend to 
>about 1.2.
>Joseph P. Noel, Ph.D.
>Investigator, Howard Hughes Medical Institute
>Professor, The Jack H. Skirball Center for Chemical Biology and Proteomics
>The Salk Institute for Biological Studies
>10010 North Torrey Pines Road
>La Jolla, CA  92037 USA
>Phone: (858) 453-4100 extension 1442
>Cell: (858) 349-4700
>Fax: (858) 597-0855
>E-mail: noel at salk.edu
>Web Site (Salk): http://www.salk.edu/faculty/faculty_details.php?id=37
>Web Site (HHMI): http://hhmi.org/research/investigators/noel.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://phenix-online.org/pipermail/phenixbb/attachments/20101219/a4d1150f/attachment-0003.htm>

More information about the phenixbb mailing list