I'm pretty sure I know the source of the degrees of freedom discrepancy. I have several other similar systems (also thermolysin, same box size, different ligands) which I am also going to simulate. They don't all have an identical number of water molecules, but the results would be easier to summarize/interpret if they were all run at the same temperatures. I assumed that since the variance in the number of water molecules/degrees of freedom was small compared to the total number that the effect of that difference on the exchange probability would be negligible. Was that a very flawed assumption?
Chris Thiebaut
Graduate Student
Purdue University
----- Original Message -----
From: "Christopher A Thiebaut" <cthieba.purdue.edu>
To: "AMBER Mailing List" <amber.ambermd.org>
Sent: Wednesday, April 9, 2014 1:20:32 PM
Subject: Re: [AMBER] Really Weird Problem with Replica Exchange
Is there any way to see the non-truncated version?
Chris Thiebaut
----- Original Message -----
From: "Jason Swails" <jason.swails.gmail.com>
To: amber.ambermd.org
Sent: Wednesday, April 9, 2014 1:20:34 PM
Subject: Re: [AMBER] Really Weird Problem with Replica Exchange
On Wed, 2014-04-09 at 13:05 -0400, Christopher A Thiebaut wrote:
> The problem is that I'm not only trying to sample the biological
> temperature ensemble. I am also interested in seeing the low
> temperature ensemble, which I cannot really do as you suggest at the
> low temperature, as the water should be frozen near their starting
> positions.
>
> After many more rounds of replica exchange, the exchange fequency are
> still unreasonably high, but have come down a little:
>
> # exchange 79
> 1 1.00 275.38-244831.05 273.71 274.96 0.81 -1
> 2 1.00 279.49-244990.76 279.98 278.73 0.89 -1
> 3 1.00 298.89-244526.61 298.19 296.86 0.91 -1
> 4 -1.00 295.95-244496.19 295.53 295.53 0.89 -1
> 5 1.00 291.07-244569.56 290.28 288.97 0.91 -1
> 6 1.00 270.99-244907.44 271.23 272.47 0.76 -1
> 7 1.00 298.69-244423.45 300.86 299.52 0.89 -1
> 8 1.00 274.76-244919.34 274.96 273.71 0.71 -1
> 9 1.00 305.26-244475.61 303.56 302.21 0.89 -1
> 10 1.00 273.72-244958.58 276.21 277.46 0.81 -1
> 11 1.00 276.11-244932.81 277.46 276.21 0.84 -1
> 12 1.00 303.57-244468.98 302.21 303.56 0.94 -1
> 13 1.00 296.16-244515.94 296.86 298.19 0.94 -1
> 14 1.00 303.38-244392.55 304.91 306.27 0.91 -1
> 15 1.00 278.82-244884.81 278.73 279.98 0.84 -1
> 16 1.00 291.47-244567.89 291.58 292.90 0.86 -1
> 17 1.00 306.35-244337.11 306.27 304.91 0.81 -1
> 18 1.00 282.77-244827.21 282.53 281.25 0.89 -1
> 19 1.00 284.39-244725.67 286.38 287.68 0.84 -1
> 20 1.00 274.02-244991.41 272.47 271.23 0.81 -1
> 21 1.00 288.43-244555.60 288.97 290.28 0.89 -1
> 22 -1.00 312.95-244137.51 310.44 310.44 0.03 -1
> 23 1.00 292.55-244691.55 292.90 291.58 0.91 -1
> 24 1.00 283.04-244674.30 283.81 285.09 0.89 -1
> 25 -1.00 270.18-244984.45 270.00 270.00 0.58 -1
> 26 1.00 306.76-244249.03 307.64 309.01 0.84 -1
> 27 1.00 284.72-244649.97 285.09 283.81 0.84 -1
> 28 1.00 297.21-244399.98 299.52 300.86 0.86 -1
> 29 -1.00 295.90-244605.15 294.21 294.21 0.86 -1
> 30 1.00 281.88-244843.69 281.25 282.53 0.89 -1
> 31 1.00 309.55-244288.13 309.01 307.64 0.56 -1
> 32 1.00 286.96-244678.21 287.68 286.38 0.91 -1
>
> Something that seems really pathological to me is that the velocity
> scaling keeps coming out to 1.00 every exchange. Shouldn't it pretty
> much never come out to exacltly 1.00?
The velocity scaling field is limited to 2 decimal places. So all 1.00
means is that it's somewhere less than 1.005. Since your temperatures
are so close together, the scaling factor is quite small.
> This seems to imply that the velocities should be the same for all
> of those temperatures accross the entire distribution, which is
> obviously not true. Has anyone very had the velocity scaling keep
> coming out at 1.00? What does that mean?
The scaling factors are not _really_ 1 exactly. You can increase the
precision of the printout in remd_exchg.F90 and recompile to satisfy
yourself here.
That said, I've never seen the temperature range from the REMD
temperature generator website be so far off before. You should look
into the source of the degrees-of-freedom discrepancy you described
earlier. The DOF has a _direct_ effect on replica spacing (it's
actually additional DOF that requires finer spacing, not additional
particles, per se).
HTH,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Apr 09 2014 - 11:00:02 PDT
