Re: Accelerated MD: question about dual boost mode

From: Jeff Wereszczynski (jmweresz_at_mccammon.ucsd.edu)
Date: Thu Mar 31 2011 - 17:07:07 CDT

Hi Patrick,

The short answer to your question is yes, the boost is too large. Remember
that the boost is applied to the potential energy of the system, which is
usually very negative. So if you have a potential energy of -50,000
kcal/mol, and you set E=10 and alpha=2, your boost will be:

(50000-10)^2/(2+(50000-10))=49,992 kcal/mol, which will make any system
unstable very fast!

In general choosing your parameters can take some trial and error. For the
amber force field we do have a rule of thumb that might be helpful. Run a
short classical MD simulation and calculate the average dihedral potential
energy (V_d) and the average total potential energy (V_tot). For a system
with Nres residues and Natom atoms, a good starting point for the dihedral
parameters may be:

E_dihed=V_d+4*Nres
alpha_dihed=4*Nres/5

And for the total you could try:

E_tot=V_tot+N_atom/5
alpha_tot= N_atom/5

Note that since V_tot should be a very negative number, E_tot should be very
negative as well. If you are not getting enough acceleration, or you are
getting too much, you might want to try incrementing E_tot and E_dihed by
their respective alpha values. Of course, these are just suggestions,
results may vary.

Also, I said this was for the amber force field. The dihedral energy is
usually larger in charmm then in amber, so I'm not sure how well those rules
will work, but they are probably an ok starting points.

Hope that helps!
Jeff

-- 
Jeff Wereszczynski
NIH Postdoctoral Fellow
University of California, San Diego
http://mccammon.ucsd.edu/~jwereszc
On Thu, Mar 31, 2011 at 1:01 PM, patrick wintrode <pat_wde2_at_yahoo.com>wrote:
> Hi.
>
> I've been testing out the new accelerated MD functionality in NAMD. I find
> that applying the boost just to the dihedrals works fine. However, if I use
> the dual boost mode, even quite small boost energies for the total-dihedral
> potential causes the simulation to become unstable and crash due to a
> constraint failure almost instantly (within ~15 steps).
>
> My system is a 372 residue protein and the total system size
> protein+solvent is ~45,000 atoms.
>
> The dihedral boost is E = 600 and alpha = 120. The smallest total-dihedral
> boost I've tried so far is E = 10 and alpha = 2.
>
> Is my total-dihedral boost just too large? From reading the McCammon
> group's papers, it doesn't appear to be.
>
> Any suggestions would be appreciated.
>
> Thanks.
>
> Patrick L. Wintrode
> Assistant Professor
> Dept. of Physiology & Biophysics
> Case Western Reserve University
> Cleveland, Ohio 44106
>

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:20:03 CST