<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
On 29/02/2012 9:26 PM, Steven Neumann wrote:
<blockquote
cite="mid:CAKZJqQEcSFZ_cDnHyHpUvmoxwGP+k5JyQwvc1X42SRtBjBrX0Q@mail.gmail.com"
type="cite"><br>
<br>
<div class="gmail_quote">On Tue, Feb 28, 2012 at 10:50 PM, Justin
A. Lemkul <span dir="ltr"><<a moz-do-not-send="true"
href="mailto:jalemkul@vt.edu">jalemkul@vt.edu</a>></span>
wrote:<br>
<blockquote style="BORDER-LEFT:#ccc 1px solid;MARGIN:0px 0px 0px
0.8ex;PADDING-LEFT:1ex" class="gmail_quote"><br>
<br>
Steven Neumann wrote:
<div>
<div class="h5"><br>
<blockquote style="BORDER-LEFT:#ccc 1px solid;MARGIN:0px
0px 0px 0.8ex;PADDING-LEFT:1ex" class="gmail_quote"><br>
<br>
On Tue, Feb 28, 2012 at 6:33 PM, Justin A. Lemkul <<a
moz-do-not-send="true" href="mailto:jalemkul@vt.edu"
target="_blank">jalemkul@vt.edu</a> <mailto:<a
moz-do-not-send="true" href="mailto:jalemkul@vt.edu"
target="_blank">jalemkul@vt.edu</a>>> wrote:<br>
<br>
<br>
<br>
Steven Neumann wrote:<br>
<br>
I run energy minimization of my protein with
implicit solvent:<br>
constraints = none<br>
<br>
integrator = steep<br>
<br>
dt = 0.001 ; ps<br>
<br>
nsteps = 30000<br>
<br>
vdwtype = cut-off<br>
<br>
coulombtype = cut-off<br>
<br>
pbc = no<br>
<br>
nstlist = 0<br>
<br>
ns_type = simple<br>
<br>
rlist = 0 ; this means all-vs-all (no
cut-off),<br>
<br>
which gets expensive for bigger systems<br>
<br>
rcoulomb = 0<br>
<br>
rvdw = 0<br>
<br>
comm-mode = angular<br>
<br>
comm-grps = Protein<br>
<br>
optimize_fft = yes<br>
<br>
;<br>
<br>
; Energy minimizing stuff<br>
<br>
;<br>
<br>
emtol =<br>
<br>
5.0<br>
<br>
emstep = 0.01<br>
<br>
;<br>
<br>
; Implicit solvent<br>
<br>
;<br>
<br>
implicit_solvent = GBSA<br>
<br>
gb_algorithm = OBC<br>
<br>
nstgbradii = 1<br>
<br>
rgbradii = 0 ; [nm] Cut-off for the
calculation of the<br>
<br>
Born radii. Currently must be equal to rlist<br>
<br>
gb_epsilon_solvent = 80 ; Dielectric constant
for the<br>
implicit solvent<br>
<br>
; gb_saltconc = 0 ; Salt concentration for
implicit solvent<br>
<br>
models, currently not used<br>
<br>
sa_algorithm = Ace-approximation<br>
<br>
sa_surface_tension = 2.25936<br>
<br>
And I obtained:<br>
Steepest Descents:<br>
<br>
Tolerance (Fmax) = 1.00000e+03<br>
<br>
Number of steps = 30000<br>
<br>
Step Time Lambda<br>
<br>
0 0.00000 0.00000<br>
<br>
  <br>
<br>
Steepest Descents converged to Fmax < 1000 in
1 steps<br>
<br>
Potential Energy = inf<br>
<br>
Maximum force = 0.0000000e+00 on atom 0<br>
<br>
Norm of force = nan<br>
<br>
Can you please explain?<br>
<br>
<br>
Not without more information.<br>
<br>
1. What is in the system? - protein without
specified box<br>
2. Which version of Gromacs is this? - Gromacs
4.5.4<br>
3. Are you using GPU or CPU architecture? - I run
it straight from<br>
the cluster on a specific node<br>
</blockquote>
<br>
</div>
</div>
Doesn't answer the question, but I'll assume CPU since EM
doesn't work on GPU's. Just checking.
<div class="im"><br>
<br>
<blockquote style="BORDER-LEFT:#ccc 1px solid;MARGIN:0px 0px
0px 0.8ex;PADDING-LEFT:1ex" class="gmail_quote"> 4. Does
an in vacuo minimization work (i.e., just turn off the GB<br>
parts)? - yes, it works<br>
<br>
</blockquote>
<br>
</div>
And what is the output?<br>
<br>
In any case, I can't reproduce any problem doing EM with
implicit solvent in 4.5.4 - my test systems work quite well.
It may just be that the system cannot be minimized because of
some clash that is unfavorable in an implicit solvent
environment (thus causing an immediate halt, though that would
be odd) but can be resolved in vacuo. Try using the structure
produced by in vacuo minimization as input into the implicit
solvent EM to see if this may be the case.
<div class="im"><br>
<br>
<blockquote style="BORDER-LEFT:#ccc 1px solid;MARGIN:0px 0px
0px 0.8ex;PADDING-LEFT:1ex" class="gmail_quote"><br>
What shoould be the value for emtol in implicit solvent
model with protein with app. 150 residues?<br>
<br>
</blockquote>
<br>
</div>
One that is sufficient for the inherent forces in the system
and for your application ;) I don't think there are any hard
and fast rules here for standard MD, and the target is not
dictated by the size of the protein.
<div class="HOEnZb">
<div class="h5"><br>
<br>
-Justin<br>
</div>
</div>
</blockquote>
<div> </div>
<div> </div>
<div>Thank you Justin.</div>
<div> </div>
<div>Mark, would you suggest equilibration e.g. for 1ns with no
restraints and no constraints in implicic solvent with
timestep =0.0005 and then run MD with implicit employing
LINCS?</div>
<div> <br>
</div>
</div>
</blockquote>
<br>
I have suggested such equilibration protocols in the past for
implicit solvent - maybe 1ns is excessive. I've had simple peptide
systems with no atomic clashes that have wanted such gentle
treatment.<br>
<br>
Mark<br>
</body>
</html>