AMBER Archive (2002)

Subject: RE: Please help. Energy suddenly increase. Protein jump out

From: Yong Duan (yduan_at_udel.edu)
Date: Wed Nov 27 2002 - 16:52:54 CST

Yes, it sure fails typically 100 or so steps after it starts conjugate
gradient minimization. If the objective is obtain a structure that has
the lowest energy, this can not do it. But if the objective is to relax
the structure a bit, to remove the bad contacts, this may work. Given
the complexity, the latter should be more realistic.
A well-minimized structure does have all bonds very close to the ideal
lengthes. But the question is how close. An average deviation of 0.001A
would give you less than 0.00025kcal/mole per bond for a bonding
constant of 500 kcal/mole/A^2. Or, 2.5 kcal/mole for a 10,000-atom
system. The "bond" energy after minimization without SHAKE is typically
much higher than this for "typical" minimizations.
Yet, 0.001A deviation would give you roughly 0.5A/ps speed (or 50
meters/sec) since our time step is 0.002ps, comparable to the typical
speed of atoms at the room temperature.
Bottom line is that you probably will see the spike even if you do an
elaborate minimization without SHAKE. But I always choose to ignore it
simply because these "velocities" (due to initial application of SHAKE
in MD) are orthogonal to the real velocities after the application of
the constraint. So, they will not propogate into the subsequent
trajectory. Your data clearly showed that. The temperature droped to
0.11K at the 3rd step from 223K. Unless you used really really short
temperature coupling constants (e.g., 0.0000001ps), I do not see how
this could be achieved by temperature regulation schemes. The only
plausible explaination is that the velocity components somehow went away
precisely because it is orthogonal to the velocities after SHAKE and
SHAKE took care of that for you. In other words, SHAKE fixed the trouble
that it gave you initially.
However, I gather your supervisor is not a specialist in this area and
it takes a few minutes to explain this, a good strategy is to use SHAKE
in minimization. Then, you need to explain why it is not such a
catastrophic event to see that energy minimization fails in the middle
of happy hours. Or, simply do not print out the energy/temperature at
every step (this sounds like that you are hiding something, though).
The third trick is a little more scholarly vigorous. You do an energy
minimization without SHAKE to convergence (or to certain number of
steps). Then, do a short (100-200 steps) energy minization with SHAKE
(in other words, stop the minimization before it fails). Now, you see,
everybody is happy, including your supervisor.

Good luck!

yong

-----Original Message-----
From: Carlos Simmerling [mailto:carlos_at_csb.sunysb.edu]
Sent: Wednesday, November 27, 2002 4:35 PM
To: Yong Duan; amber_at_heimdal.compchem.ucsf.edu; 'Lich Nguyen'
Subject: Re: Please help. Energy suddenly increase. Protein jump out

Be aware that SHAKE can often fail during minimization.
A well-minimized structure should have bond lengths
very close to the reference values, unless there is some
other problem with the conformation.
Carlos

----- Original Message -----
From: "Yong Duan" <yduan_at_udel.edu>
To: <amber_at_heimdal.compchem.ucsf.edu>; "'Lich Nguyen'"
<nguyenthanhlich_at_yahoo.com>
Sent: Wednesday, November 27, 2002 1:42 PM
Subject: RE: Please help. Energy suddenly increase. Protein jump out

>
> My guess is that you did not apply SHAKE when you minimized the
> structure. If this is the case, I think you do not have to worry about

> it. But if you are concerned, apply the same level of SHAKE in
> minimization as the one that you will use in MD. The spike will
> magically disappear (in most cases).
>
> yong
>
> -----Original Message-----
> From: David A. Case [mailto:case_at_scripps.edu]
> Sent: Wednesday, November 27, 2002 11:27 AM
> To: Lich Nguyen
> Cc: amber_at_heimdal.compchem.ucsf.edu
> Subject: Re: Please help. Energy suddenly increase. Protein jump out
>
>
> On Tue, Nov 26, 2002, Lich Nguyen wrote:
> >
> > However, my supervisor wonders something must be wrong because my
> > protein jump out at the first time step of simulation and both EP
> > and EK increase suddenly.
> > Time step 0: EK=0 EP=-240528 Temp=0
> > 1: EK=27062 EP=-240528 Temp=223.11
> > 2: EK=22866 EP=-209771 Temp=188.52
> > 3: EK=13.9 EP=-209782 Temp=0.11
>
> You are getting an enormous spike in KE on step 1. Are you sure that
> your system is really minimized? Does the minimzation end at the same

> potential energy has you start with the dynamics?
>
> I note that your vdW energy on step zero is 32,000....this is not
> consistent with having a minimized structure. Also, the EEL energy
> (-300000) is very large...you should probably check to make sure that
> you don't have two atoms very near to each other. For example, use
> the "checkoverlap" command in ptraj to see if you have any obvious
> short distances.
>
> >
> > &wt type='TEMP0', istep1=0,istep2=1000,value1=0.0,
> > value2=5.0, &end
>
> Try just doing a "simple" (nmropt=0) short run with temp0=5.0. There
> may be a problem with having temp0=0 on the first step.
>
> ..hope this helps...dac
>
> --
>
> ==================================================================
> David A. Case | e-mail: case_at_scripps.edu
> Dept. of Molecular Biology, TPC15 | fax: +1-858-784-8896
> The Scripps Research Institute | phone: +1-858-784-9768
> 10550 N. Torrey Pines Rd. | home page:
> La Jolla CA 92037 USA | http://www.scripps.edu/case
> ==================================================================
>
>