Happy Holidays, Amber!
I have a present to put under the code tree, a system whereby one can force
two atoms within a system to come together but only at some point along a
specified path. Picture two particles confined to a rigid track, which are
being pulled towards one another by a string that also runs along the
track. Each particle connects to an atom of the real system by some short,
stiff harmonic potential. In this manner, the particles being pulled
together along the track will pull the two real atoms to which they are
connected into proximity. (The strength of the pulling is just details, but
it is like a distance-based NMR restraint and strong enough to pull the
particles together.)
Now, the question: can I still use conjugate gradient minimization to guide
this system to its minimum? The way to compute forces seems to be:
1.) Compute forces due to all real atoms on all others.
2.) Compute forces of dummy particles pulling on their real atom
connections.
3.) Compute forces of dummy particles on one another.
4.) Sum results of steps 1-3 and remove the rejection of forces acting on
dummy particle forces relative to the path's direction where the dummy
particles sit (that is, the force on each dummy particle is reduced to the
force that strictly moves in the direction of the path.
5.) Compute the conjugate gradient.
6.) Again remove the rejection of the conjugate gradient so that forces on
the dummy particles specifically push in the direction of the path.
7.) Move real atoms incrementally in x, y, and z, and dummy particles
equivalent distances along the path.
8.) Return to step 1, with whatever memory is held in the CG data vectors,
cycle until complete.
See, this looks just like regular CG except for step 6, because I am not
going to take those particles off the path, even incrementally. Is this a
sound way to proceed?
Thanks,
Dave
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Tue Dec 27 2016 - 15:30:02 PST
