Re: [gmx-users] problems with the output of pullx

[Alfredo E. Cardenas]

Hi all,
I want to update my own post for any user that could have similar issues in the 
future. The issue that I described before (large spikes were observed in the 
values reported in pullx file but when i calculated the same restrained 
distance using “gmx distance” no such spikes were observed). The problem was 
the pbcatom (reference atom for the treatment of PBC inside a group). I thought 
I didn’t have that problem because I was only restraining the z coordinate and 
the peptide I was pulling inside the membrane never reach near the walls in the 
z direction. The problem was with the pbcatom chosen for the membrane group. 
The pbcatom that was chosen by gromacs was a hydrogen in the choline group 
region and that certainly increases the possibility that some lipids in the 
other layer move to the wrong side of the box and create havoc during the 
pulling calculation. Once I explicitly assigned a different pbcatom in the mdp 
(for example, the terminal methyl carbon of one of the lipids), the spikes in 
the pullx file don’t show up anymore.
By the way, the problem was described in an earlier post:
https://mailman-1.sys.kth.se/pipermail/gromacs.org_gmx-developers/2010-April/004198.html
 


Thanks,
Alfredo



> On Feb 24, 2018, at 4:43 PM, alfredo  wrote:
> 
> Hi Mark,
> 
> Thanks for your comment. No, that is not the problem. At that location the 
> center of mass of the peptide is deep inside the membrane, separation between 
> the two pulling groups is 0.4 nm and the dimension of the cell along z is 
> more than 10 nm. I am only pulling along the z direction. The puzzle to me is 
> that when I extract the center of mass separation along z between the same 
> two groups using gmx traj those spikes don't show up at the times when they 
> are shown in the pullx file.
> 
> Alfredo
> 
> 
> 
> On 2018-02-24 11:57, Mark Abraham wrote:
>> Hi,
>> My (thoroughly uneducated) guess is that the spikes are related to the pull
>> distance approaching half of the dimensions of the cell. Not all flavours
>> of pulling can  handle this. Might that be the issue?
>> Mark
>> On Sat, Feb 24, 2018, 17:55 alfredo  wrote:
>>> Hi,
>>> Updating my post. The problem has been observed in two different machine
>>> systems (the latest I have found the problem was the skylake nodes in
>>> tacc). I assumed it has to be some communication bug of coordinates and
>>> forces in the pull part of the code. Probably observed in my case
>>> because of the large size of the peptide I am pulling inside the
>>> membrane. For now I am thinking to extract coordinates from the trr file
>>> and from them compute the pulling harmonic forces. But not an ideal
>>> solution.
>>> Thanks
>>> Alfredo
>>> On 2018-02-22 10:17, Alfredo E. Cardenas wrote:
>>> > Hi,
>>> > I am using gromacs to get the PMF of a peptide of  about 20 amino
>>> > acids, moving inside of a bilayer membrane. After pulling the peptide
>>> > inside the membrane now I am using  pull-coord1-type = umbrella
>>> > and pull-coord1-geometry  = distance to sample configurations in each
>>> > window for the umbrella simulations along the z axis (axis
>>> > perpendicular to the membrane surface). Runs finish ok, no error
>>> > messages. The problem is that when I looked at the contents of the
>>> > pullx file I observed spikes (up to 5 or more Angstroms) in the z
>>> > coordinate separating the center of mass of the peptide from the
>>> > membrane center. But when I extract the z coordinates of the center of
>>> > mass of the two groups and compute the difference the values look
>>> > reasonable with no spikes.
>>> >
>>> > Here an example (it starts good):
>>> >   time (ps)   from pullx  from traj analysis
>>> >
>>> >20.000  0.475923002  0.475919992
>>> >200010.000  0.498394012  0.498389989
>>> >200020.000  0.527589977  0.527589977
>>> >200030.000  0.491834015  0.493739992
>>> >200040.000  0.485377997  0.485379994
>>> >200050.000  0.488474995  0.488469988
>>> >200060.000  0.507991016  0.507990003
>>> >200070.000  0.475095987  0.475100011
>>> >200080.000  0.465889990  0.465889990
>>> >200090.000  0.515878975  0.515879989
>>> >200100.000  0.501435995  0.501429975
>>> >200110.000  0.505191982  0.505190015
>>> >
>>> > Here a bad section:
>>> >
>>> >214000.000  0.427343011  0.601450026
>>> >214010.000  0.484564990  0.545799971
>>> >214020.000  0.530139029  0.603110015
>>> >214030.000  0.176231995  0.650319993
>>> >214040.000  0.342045009  0.637109995
>>> >214050.000  0.181202993  0.636659980
>>> >214060.000  0.338808000  0.595300019
>>> >214070.000  0.442301005  0.547529995
>>> >

Re: [gmx-users] problems with the output of pullx

[alfredo]

Hi Mark,

Thanks for your comment. No, that is not the problem. At that location 
the center of mass of the peptide is deep inside the membrane, 
separation between the two pulling groups is 0.4 nm and the dimension of 
the cell along z is more than 10 nm. I am only pulling along the z 
direction. The puzzle to me is that when I extract the center of mass 
separation along z between the same two groups using gmx traj those 
spikes don't show up at the times when they are shown in the pullx file.


Alfredo



On 2018-02-24 11:57, Mark Abraham wrote:

Hi,

My (thoroughly uneducated) guess is that the spikes are related to the 
pull
distance approaching half of the dimensions of the cell. Not all 
flavours

of pulling can  handle this. Might that be the issue?

Mark

On Sat, Feb 24, 2018, 17:55 alfredo  wrote:


Hi,
Updating my post. The problem has been observed in two different 
machine

systems (the latest I have found the problem was the skylake nodes in
tacc). I assumed it has to be some communication bug of coordinates 
and

forces in the pull part of the code. Probably observed in my case
because of the large size of the peptide I am pulling inside the
membrane. For now I am thinking to extract coordinates from the trr 
file

and from them compute the pulling harmonic forces. But not an ideal
solution.
Thanks
Alfredo




On 2018-02-22 10:17, Alfredo E. Cardenas wrote:
> Hi,
> I am using gromacs to get the PMF of a peptide of  about 20 amino
> acids, moving inside of a bilayer membrane. After pulling the peptide
> inside the membrane now I am using  pull-coord1-type = umbrella
> and pull-coord1-geometry  = distance to sample configurations in each
> window for the umbrella simulations along the z axis (axis
> perpendicular to the membrane surface). Runs finish ok, no error
> messages. The problem is that when I looked at the contents of the
> pullx file I observed spikes (up to 5 or more Angstroms) in the z
> coordinate separating the center of mass of the peptide from the
> membrane center. But when I extract the z coordinates of the center of
> mass of the two groups and compute the difference the values look
> reasonable with no spikes.
>
> Here an example (it starts good):
>   time (ps)   from pullx  from traj analysis
>
>20.000  0.475923002  0.475919992
>200010.000  0.498394012  0.498389989
>200020.000  0.527589977  0.527589977
>200030.000  0.491834015  0.493739992
>200040.000  0.485377997  0.485379994
>200050.000  0.488474995  0.488469988
>200060.000  0.507991016  0.507990003
>200070.000  0.475095987  0.475100011
>200080.000  0.465889990  0.465889990
>200090.000  0.515878975  0.515879989
>200100.000  0.501435995  0.501429975
>200110.000  0.505191982  0.505190015
>
> Here a bad section:
>
>214000.000  0.427343011  0.601450026
>214010.000  0.484564990  0.545799971
>214020.000  0.530139029  0.603110015
>214030.000  0.176231995  0.650319993
>214040.000  0.342045009  0.637109995
>214050.000  0.181202993  0.636659980
>214060.000  0.338808000  0.595300019
>214070.000  0.442301005  0.547529995
>214080.000  0.396046013  0.565050006
>214090.000  0.431407988  0.538460016
>214100.000  0.402586013  0.56825
>214110.000  0.438223004  0.575810015
>
> Then good again:
>
>23.000  0.477869004  0.477869987
>230010.000  0.511840999  0.511839986
>230020.000  0.469146013  0.469150007
>230030.000  0.480194002  0.480190009
>230040.000  0.525618017  0.525619984
>230050.000  0.498955995  0.498950005
>230060.000  0.489230990  0.489230007
>230070.000  0.531931996  0.531930029
>230080.000  0.535376012  0.535380006
>230090.000  0.488822013  0.48883
>230100.000  0.510704994  0.510699987
>230110.000  0.524537981  0.524540007
>230120.000  0.513199985  0.513189971
>
> This problem happens in most umbrella windows that I have examined,
> sometimes several times during the long trajectories I am running. The
> pullf output also have those spikes.
>
> Here is the mdp file I am using:
>
> integrator  = md
> dt  = 0.002
> nsteps  = 5000
> nstlog  = 1
> nstxout = 5000
> nstvout = 5000
> nstfout = 5000
> nstcalcenergy   = 500
> nstenergy   = 500
> ;
> cutoff-scheme   = Verlet
> nstlist = 20
> rlist   = 1.2
> coulombtype = pme
> rcoulomb= 1.2
> vdwtype = Cut-off
> vdw-modifier= Force-switch
> 

Re: [gmx-users] problems with the output of pullx

[Mark Abraham]

Hi,

My (thoroughly uneducated) guess is that the spikes are related to the pull
distance approaching half of the dimensions of the cell. Not all flavours
of pulling can  handle this. Might that be the issue?

Mark

On Sat, Feb 24, 2018, 17:55 alfredo  wrote:

> Hi,
> Updating my post. The problem has been observed in two different machine
> systems (the latest I have found the problem was the skylake nodes in
> tacc). I assumed it has to be some communication bug of coordinates and
> forces in the pull part of the code. Probably observed in my case
> because of the large size of the peptide I am pulling inside the
> membrane. For now I am thinking to extract coordinates from the trr file
> and from them compute the pulling harmonic forces. But not an ideal
> solution.
> Thanks
> Alfredo
>
>
>
>
> On 2018-02-22 10:17, Alfredo E. Cardenas wrote:
> > Hi,
> > I am using gromacs to get the PMF of a peptide of  about 20 amino
> > acids, moving inside of a bilayer membrane. After pulling the peptide
> > inside the membrane now I am using  pull-coord1-type = umbrella
> > and pull-coord1-geometry  = distance to sample configurations in each
> > window for the umbrella simulations along the z axis (axis
> > perpendicular to the membrane surface). Runs finish ok, no error
> > messages. The problem is that when I looked at the contents of the
> > pullx file I observed spikes (up to 5 or more Angstroms) in the z
> > coordinate separating the center of mass of the peptide from the
> > membrane center. But when I extract the z coordinates of the center of
> > mass of the two groups and compute the difference the values look
> > reasonable with no spikes.
> >
> > Here an example (it starts good):
> >   time (ps)   from pullx  from traj analysis
> >
> >20.000  0.475923002  0.475919992
> >200010.000  0.498394012  0.498389989
> >200020.000  0.527589977  0.527589977
> >200030.000  0.491834015  0.493739992
> >200040.000  0.485377997  0.485379994
> >200050.000  0.488474995  0.488469988
> >200060.000  0.507991016  0.507990003
> >200070.000  0.475095987  0.475100011
> >200080.000  0.465889990  0.465889990
> >200090.000  0.515878975  0.515879989
> >200100.000  0.501435995  0.501429975
> >200110.000  0.505191982  0.505190015
> >
> > Here a bad section:
> >
> >214000.000  0.427343011  0.601450026
> >214010.000  0.484564990  0.545799971
> >214020.000  0.530139029  0.603110015
> >214030.000  0.176231995  0.650319993
> >214040.000  0.342045009  0.637109995
> >214050.000  0.181202993  0.636659980
> >214060.000  0.338808000  0.595300019
> >214070.000  0.442301005  0.547529995
> >214080.000  0.396046013  0.565050006
> >214090.000  0.431407988  0.538460016
> >214100.000  0.402586013  0.56825
> >214110.000  0.438223004  0.575810015
> >
> > Then good again:
> >
> >23.000  0.477869004  0.477869987
> >230010.000  0.511840999  0.511839986
> >230020.000  0.469146013  0.469150007
> >230030.000  0.480194002  0.480190009
> >230040.000  0.525618017  0.525619984
> >230050.000  0.498955995  0.498950005
> >230060.000  0.489230990  0.489230007
> >230070.000  0.531931996  0.531930029
> >230080.000  0.535376012  0.535380006
> >230090.000  0.488822013  0.48883
> >230100.000  0.510704994  0.510699987
> >230110.000  0.524537981  0.524540007
> >230120.000  0.513199985  0.513189971
> >
> > This problem happens in most umbrella windows that I have examined,
> > sometimes several times during the long trajectories I am running. The
> > pullf output also have those spikes.
> >
> > Here is the mdp file I am using:
> >
> > integrator  = md
> > dt  = 0.002
> > nsteps  = 5000
> > nstlog  = 1
> > nstxout = 5000
> > nstvout = 5000
> > nstfout = 5000
> > nstcalcenergy   = 500
> > nstenergy   = 500
> > ;
> > cutoff-scheme   = Verlet
> > nstlist = 20
> > rlist   = 1.2
> > coulombtype = pme
> > rcoulomb= 1.2
> > vdwtype = Cut-off
> > vdw-modifier= Force-switch
> > rvdw_switch = 1.0
> > rvdw= 1.2
> > ;
> > tcoupl  = Nose-Hoover
> > tc_grps = PROT   MEMB   SOL_ION
> > tau_t   = 1.01.01.0
> > ref_t   = 303.15 303.15 303.15
> > ;
> > pcoupl  = Parrinello-Rahman
> > pcoupltype  = semiisotropic
> > tau_p  

Re: [gmx-users] problems with the output of pullx

[alfredo]

Hi,
Updating my post. The problem has been observed in two different machine 
systems (the latest I have found the problem was the skylake nodes in 
tacc). I assumed it has to be some communication bug of coordinates and 
forces in the pull part of the code. Probably observed in my case 
because of the large size of the peptide I am pulling inside the 
membrane. For now I am thinking to extract coordinates from the trr file 
and from them compute the pulling harmonic forces. But not an ideal 
solution.

Thanks
Alfredo




On 2018-02-22 10:17, Alfredo E. Cardenas wrote:

Hi,
I am using gromacs to get the PMF of a peptide of  about 20 amino
acids, moving inside of a bilayer membrane. After pulling the peptide
inside the membrane now I am using  pull-coord1-type = umbrella
and pull-coord1-geometry  = distance to sample configurations in each
window for the umbrella simulations along the z axis (axis
perpendicular to the membrane surface). Runs finish ok, no error
messages. The problem is that when I looked at the contents of the
pullx file I observed spikes (up to 5 or more Angstroms) in the z
coordinate separating the center of mass of the peptide from the
membrane center. But when I extract the z coordinates of the center of
mass of the two groups and compute the difference the values look
reasonable with no spikes.

Here an example (it starts good):
time (ps)   from pullx  from traj analysis

   20.000  0.475923002  0.475919992
   200010.000  0.498394012  0.498389989
   200020.000  0.527589977  0.527589977
   200030.000  0.491834015  0.493739992
   200040.000  0.485377997  0.485379994
   200050.000  0.488474995  0.488469988
   200060.000  0.507991016  0.507990003
   200070.000  0.475095987  0.475100011
   200080.000  0.465889990  0.465889990
   200090.000  0.515878975  0.515879989
   200100.000  0.501435995  0.501429975
   200110.000  0.505191982  0.505190015

Here a bad section:

   214000.000  0.427343011  0.601450026
   214010.000  0.484564990  0.545799971
   214020.000  0.530139029  0.603110015
   214030.000  0.176231995  0.650319993
   214040.000  0.342045009  0.637109995
   214050.000  0.181202993  0.636659980
   214060.000  0.338808000  0.595300019
   214070.000  0.442301005  0.547529995
   214080.000  0.396046013  0.565050006
   214090.000  0.431407988  0.538460016
   214100.000  0.402586013  0.56825
   214110.000  0.438223004  0.575810015

Then good again:

   23.000  0.477869004  0.477869987
   230010.000  0.511840999  0.511839986
   230020.000  0.469146013  0.469150007
   230030.000  0.480194002  0.480190009
   230040.000  0.525618017  0.525619984
   230050.000  0.498955995  0.498950005
   230060.000  0.489230990  0.489230007
   230070.000  0.531931996  0.531930029
   230080.000  0.535376012  0.535380006
   230090.000  0.488822013  0.48883
   230100.000  0.510704994  0.510699987
   230110.000  0.524537981  0.524540007
   230120.000  0.513199985  0.513189971

This problem happens in most umbrella windows that I have examined,
sometimes several times during the long trajectories I am running. The
pullf output also have those spikes.

Here is the mdp file I am using:

integrator  = md
dt  = 0.002
nsteps  = 5000
nstlog  = 1
nstxout = 5000
nstvout = 5000
nstfout = 5000
nstcalcenergy   = 500
nstenergy   = 500
;
cutoff-scheme   = Verlet
nstlist = 20
rlist   = 1.2
coulombtype = pme
rcoulomb= 1.2
vdwtype = Cut-off
vdw-modifier= Force-switch
rvdw_switch = 1.0
rvdw= 1.2
;
tcoupl  = Nose-Hoover
tc_grps = PROT   MEMB   SOL_ION
tau_t   = 1.01.01.0
ref_t   = 303.15 303.15 303.15
;
pcoupl  = Parrinello-Rahman
pcoupltype  = semiisotropic
tau_p   = 5.0
compressibility = 4.5e-5  4.5e-5
ref_p   = 1.0 1.0
;
constraints = h-bonds
constraint_algorithm= LINCS
continuation= yes
;
nstcomm = 500
comm_mode   = linear
comm_grps   = PROT_MEMB   SOL_ION
;
refcoord_scaling= com
;
pull= yes
pull-coord1-type= umbrella
pull-coord1-geometry= distance
pull-coord1-dim = N N Y
pull-ngroups= 2
pull-ncoords= 1
pull-coord1-groups  = 1 2
pull-group1-name= MEMB
pull-group2-name= PROT
pull-coord1-init= 0.400
pull-coord1-k