OK. Two things:
[1] I just checked the structure output from the problematic calculation (Amber10; sander.MPI; bugfixes applied through #35). The input structure was folded as expected -- however, the minimized AMBER output structure is now unfolded.
Here is the information for the first distance restraint:
# 49 VAL HA 52 LYS H 4.11 (#peak 13 #SUP 1.00 #QF 0.99)
&rst
ixpk= 13, nxpk= 0, iat= 810, 868, r1= 1.30, r2= 1.80, r3= 4.11, r4= 4.61,
rk2=20.0, rk3=20.0, ir6=1, ialtd=0,
&end
Information concerning this restraint in the output:
------------------------------------------------------------------------------
First atom Last atom curr. value target deviation penalty
------------------------------------------------------------------------------
HA VAL 49 -- H LYS 52: 3.143 4.610 1.467 0.000 d 13: 0
It's almost as if the distance restraints are not being applied during the calculation, allowing the input structure to unfold during the minimization.
[2] I ran a calculation using the exact same data as above (problematic results) on another cluster (Amber10; sander.MPI; bugfixes applied through #17). This gives results as expected:
NSTEP ENERGY RMS GMAX NAME NUMBER
100 -3.9595E+03 1.1294E+00 1.4438E+01 CZ3 840
BOND = 40.6235 ANGLE = 95.1117 DIHED = 736.2988
VDWAALS = -436.9454 EEL = -3020.7003 EGB = -1996.4825
1-4 VDW = 276.2185 1-4 EEL = 316.0835 RESTRAINT = 30.3154
EAMBER = -3989.7922
NMR restraints: Bond = 1.435 Angle = 0.000 Torsion = 0.382
===============================================================================
Both the input and minimized AMBER output structures are folded (as expected).
The information from [2] suggests to me there is nothing wrong with my input files, and there is something wrong with my sander.MPI binary (bugfixes applied through #35). I will run with LISTIN=POUT and send the results (if you still think that will help to debug).
Doug
On May 21, 2010, at 3:01 PM, case wrote:
> On Fri, May 21, 2010, Douglas Kojetin wrote:
>
>> Setting pencut = -1.0, here are the first few distance restraints:
>>
>> ---------------------------------------------------------------------------
>> ---
>> First atom Last atom curr. value target deviation penalty
>> ---------------------------------------------------------------------------
>> ---
>> HA VAL 49 -- H LYS 52: 3.143 4.610 1.467 0.000 d
>> 13: 0
>
> OK...now we have progress; you can examine the input data in the RST file,
> looking just at the first restraint. If the solution is not obvious, you
> could post just that one piece of input and ask questions. Also, check
> your mdin file to make sure you have not set some restraint weight to
> zero.
>
> Not to beat a dead horse, but my previous suggestion of setting LISTIN=POUT
> still seems like a good idea -- it gives lots more detail about what is going
> on.
>
> I think it is rather unlikely that this is related to serial vs. parallel, or
> would be fixed by some recompilation. There are lots of ways that restraints
> can do things you don't expect, but difficult to remotely debug without more
> information. (We would need to know r1,r2,r3,r4, rk2, rk3 at a minimum.)
>
> ....dac
>
>
> _______________________________________________
> 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 Sat May 22 2010 - 10:30:03 PDT
