Hi,
I am interested in running an MD simulation with a protein-ligand system.
Nonetheless, I keep running into a memory access error in the first MD
computation (after minimization) which I can't figure out how to solve. To
simplify everything, I ran an MD simulation of the protein alone and
everything worked smoothly. With the ligand alone, it keeps returning this
error though.
Ligand parameterization details: I optimized the ligand (squalene molecule)
geometry with Gaussian (options:
#opt b3lyp/6-31+g geom=connectivity). Next, I computed the resp charges
also with gaussian (options: #p hf/6-31+g geom=check scf=tight pop=mk
iop(6/33=2)). Finally, I generated the prepi file using antechamber
(antechamber -i res-charge.log -fi gout -o SQL.gaussian.prepi -fo prepi -c
resp -nc -1 -m 1 -rn SQL -at amber ) and the parameters file with parmchk2
( parmchk2 -i SQL.gaussian.prepi -f prepi -o SQL.gaff.frcmod ).
Box simulation: I generated x_solv.prmtop and x_solv.ncrst with tleap
containing the ligand alone with waters and Na+ Cl- ions. These files are
attached for reference.
Minimization (min-1.mdin): 2500 steepest descent minimization followed by
2500 conjugate gradient. First 2500 steps work great. When it switches to
CG, the algorithm starts returning ".... RESTARTED DUE TO LINMIN FAILURE
..." and it stops after a couple more then 200 steps. I read this should be
a problem itself, so I proceed.
MD simulation in the NVT ensemble (nvt.mdmin). No error is reported in the
output. In the command line, "an illegal memory access was encountered
launching kernel kClearForces" is reported.
Additional details:
1) I'm running the simulations with AMBER20 pmemd.cuda_SPFP @GPUK40 and
.GPUK20 machines.
2) I have already run MD simulation with other ligands and it worked
perfectly using this same strategy.
I attached the files here (
https://1drv.ms/u/s!AqJBrTd9Fp4I2nxFkJP7xwPNzvaU?e=e7VBuJ) which should be
enough to reproduce my simulation if desired.
Let me know if you need additional information.
Looking forward to hearing from you.
Best regards,
--
*Állan*
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Oct 13 2020 - 07:30:02 PDT
