[gmx-users] Re: openmm

[Alan]

Thanks Justin.

I could swear I tried what you said... anyway it worked now (2.5 x
faster), but not with implicit solvent for mdrun-openmm... Trying with
plain mdrun with the setup for implicit solvent and it seemed to work
(although I don't know how to interpret this from
GromacsOpenMMPreview3-Mac/Gromacs-OpenMM Readme.txt file: "even though
this is not available in the standard Gromacs 4")

I still would like an example/tutorial that really worked for someone
with gmx openmm.

In the end, with mdrun-openmm and implicit solvent in the mdp file, I
got a file md.gro full of "nan     nan".

For those interested, here are the commands I did. If you have GMX
with ffamber, CUDA and GMX-openmm you may reproduce it.

###############################

cat << EOF >| em.mdp
define                   = -DFLEXIBLE
integrator               = cg ; steep
nsteps                   = 200
constraints              = none
emtol                    = 1000.0
nstcgsteep               = 10 ; do a steep every 10 steps of cg
emstep                   = 0.01 ; used with steep
nstcomm                  = 1
coulombtype              = Reaction-Field
ns_type                  = simple
rlist                    = 1.2
rcoulomb                 = 1.2
rvdw                     = 1.2
vdwtype                  = cut-off
Tcoupl                   = no
Pcoupl                   = no
gen_vel                  = no
nstxout                  = 0 ; write coords every # step
epsilon_rf               = 0
pbc = no
EOF

cat << EOF >| md.mdp
integrator               = md
nsteps                   = 1000
dt                       = 0.002
constraints              = all-bonds
nstcomm                  = 1
ns_type                  = simple
rlist                    = 1.2
rcoulomb                 = 1.2
rvdw                     = 1.2
vdwtype                  = cut-off
coulombtype              = Reaction-Field
epsilon_rf               = 0
Tcoupl                   = no
Pcoupl                   = no
gen_vel                  = yes
nstxout                  = 2 ; write coords every # step
lincs-iter               = 2
pbc = no
comm_mode = ANGULAR
; uncomment below to test with implicit solvent
;implicit_solvent = GBSA
;gb_algorithm = OBC
;gb_epsilon_solvent = 78.3
EOF

wget -c "http://www.pdb.org/pdb/download/downloadFile.do?fileFormat=pdb&compression=NO&structureId=1BVG"
-O 1BVG.pdb
grep 'ATOM  ' 1BVG.pdb>| Protein.pdb
grep 'HETATM' 1BVG.pdb>| Ligand.pdb

sed s/PRO\ A\ \ \ 1/NPROA\ \ \ 1/g Protein.pdb | sed s/PRO\ B\ \ \
1/NPROB\ \ \ 1/g \
| sed s/PHE\ A\ \ 99/CPHEA\ \ 99/g | sed s/PHE\ B\ \ 99/CPHEB\ \ 99/g \
| sed s/O\ \ \ CPHE/OC1\ CPHE/g | sed s/OXT\ CPHE/OC2\ CPHE/g \
| sed s/HIS\ /HID\ /g | sed s/LYS\ /LYP\ /g | sed s/CYS\ /CYN\ /g >|
ProteinAmber.pdb

# Process with pdb2gmx and define water
pdb2gmx -ff amber99sb -f ProteinAmber.pdb -o Protein2.pdb -p Protein.top -ignh

grompp -f em.mdp -c Protein2.pdb -p Protein.top -o em.tpr
mdrun -v -deffnm em

grompp -f md.mdp -c em.gro -p Protein.top -o md.tpr
mdrun -v -deffnm md

vmd md.gro md.trr # OK

mdrun-openmm -v -deffnm md
# 3 times fater
# works fine without set for implicit solvent in md.mdp
# with implicit solvent, md.gro full of nan nan nan everywhere.

###############################

On Sat, Jul 18, 2009 at 17:31, Alan<alanwilter at gmail.com> wrote:
> Ok, I tried.
>
> Can someone tell how to have in a mdp file this combination (in order
> to satisfy gromacs openmm) without raising a error in grompp?
>
> ns_type = simple
> pbc = no
> coulombtype = Reaction-Field
>
> Whatever I try I get:
>
> ERROR: Twin-range neighbour searching (NS) with simple NS algorithm
> not implemented
>
> My full md.mdp file is:
> integrator               = md
> nsteps                   = 1000
> dt                       = 0.002
> constraints              = all-bonds
> nstcomm                  = 1
> ns_type                  = simple
> rlist                    = 1.0
> rcoulomb                 = 1.2
> rvdw                     = 1.2
> vdwtype                  = cut-off
> coulombtype              = Reaction-Field
> epsilon_rf               = 0
> Tcoupl                   = no
> Pcoupl                   = no
> gen_vel                  = yes
> nstxout                  = 2 ; write coords every # step
> lincs-iter               = 2
> pbc = no
> implicit_solvent = GBSA
> gb_algorithm = OBC
> gb_epsilon_solvent = 78.3
> comm_mode = ANGULAR
>
>
> Many thanks in advance,
>
> Alan
>
>
>
> On Sat, Jul 18, 2009 at 11:35, Alan<alanwilter at gmail.com> wrote:
>> Dear Mark,
>>
>> Thanks. I reread those file carefully and noticed the "only implicit
>> solvent". Sorry for that. I am building my test case here to see by
>> myself how fast it can be.
>>
>> Alan
>>
>>
>> On Sat, Jul 18, 2009 at 11:00, <gmx-users-request at gromacs.org> wrote:
>>
>>> Alan wrote:
>>>> Hi list, does anyone have an example (input pdb, gmx commands and
>>>> md.mdp for example) to test gromacs with and without openmm?
>>>>
>>>> The case I use here (with explicit water) didn't show me any speed up
>>>> (comparing with mpirun -c 2 mdrun_mpi...).
>>>>
>>>> I am using Mac Leopard 10.5.7 with Cuda drivers and gromacs 4.0.5
>>>
>>> This would be expected if you read the OpenMM README.... or see
>>> http://oldwiki.gromacs.org/index.php/OpenMM_GROMACS
>>>
>>> Mark
>> --
>> Alan Wilter S. da Silva, D.Sc. - CCPN Research Associate
>> Department of Biochemistry, University of Cambridge.
>> 80 Tennis Court Road, Cambridge CB2 1GA, UK.
>>>>http://www.bio.cam.ac.uk/~awd28<<
>>
>
>
>
> --
> Alan Wilter S. da Silva, D.Sc. - CCPN Research Associate
> Department of Biochemistry, University of Cambridge.
> 80 Tennis Court Road, Cambridge CB2 1GA, UK.
>>>http://www.bio.cam.ac.uk/~awd28<<
>



-- 
Alan Wilter S. da Silva, D.Sc. - CCPN Research Associate
Department of Biochemistry, University of Cambridge.
80 Tennis Court Road, Cambridge CB2 1GA, UK.
>>http://www.bio.cam.ac.uk/~awd28<<