Re: [gmx-users] further on: very strange phenomenon for my production MD by gromacs

[Mark Abraham]

Hi,

On Fri, Apr 15, 2016 at 3:44 PM Brett  wrote:

> Dear Mark,
>
>
>
> My protein was a 6-identical subunit protein complex, initially each
> subunit contains a ligand. I want to see the conformational change from
> ligand bond to ligand free state, during this process there were reports
> that significant conformation change would occur
>
>
> First from the initial PDB with the ligands, I delete the ligands and get
> the apo pdb, after pdb2gmx etc, I process the energy minimization, nvt
> equilibration and npt equilibration, then I start the production md. During
> the initial 6 ns production md, there were some conformation change
> observed, but I do not know whether it diffuse.
>

You also have a large number of water molecules. The outermost extent of
their shape (their "envelope") was what you observed as becoming more
spherical, once you required their representation to avoid jumps over the
duration of the trajectory.

In addition, I do not understand what is "the envelope of the diffusion",


If you do a large number of one-dimensional random walks, you have a
characteristic distribution of the results, right? Here, your water
molecules are doing a three-dimensional random walk, without jumps. The
distribution of their final displacements isn't going to resemble your
starting shape. But if you choose a representation where they jump, then
the starting shape will remain.

and does diffusion leads the box from cubic to sphere?
>

Don't confuse the representation with what happened. You told mdrun to do a
periodic simulation. If you later require the representation of the
trajectory to not have jumps over boundaries, you aren't changing the
simulation.

Mark

Brett
>
>
>
>
>
>
>
>
>
> At 2016-04-15 21:30:12, "Mark Abraham"  wrote:
> >Hi,
> >
> >On Fri, Apr 15, 2016 at 3:18 PM Brett  wrote:
> >
> >> Dear All,
> >>
> >> By xmgrace I have checked the xmgrace, I find the spike was between
> >> -8.245e+06 and -8.225e+06, is the spike massive or not?
> >>
> >>
> >> In addition, by
> >> gmx trjconv -f ma_0_1.trr -s md_0_1.tpr -o 6000pswithwater.pdb --pbc
> >> nojump -dump 6000,
> >>
> >>
> >> I got the 6 ns with water pdb and I have checked the pdb (with water
> box),
> >> I find the resi 366-367 were not disconnected!
> >>
> >>
> >> However, by gmx editconf, the original water box was in cubic, after
> "gmx
> >> trjconv -f ma_0_1.trr -s md_0_1.tpr -o 6000pswithwater.pdb --pbc nojump
> >> -dump 6000", the cubic water box was replaced by a rough sphere water
> box
> >> with the protein roughly in the center. Will you please explain why the
> >> cubic water box was replaced by the round sphere water box? Anything
> wrong
> >> I have made?
> >>
> >
> >You've simulated a bunch of molecules that diffuse. What shape would you
> >expect for the envelope of the diffusion? How would it vary over time?
> >
> >Further, by visual check if in a production MD I did not see residue
> >> disconnection caused by "Periodic_Boundary_Conditions", can we
> guarantee no
> >> minor disconnection occurred during the md (or significant bond length
> >> variance unobservable by the naked eye with pymol) if we did not process
> >> the trjconv correction step?
> >>
> >
> >The forces produced by such a disconnection would blow everything up
> >immediately. Since it didn't...
> >
> >Mark
> >
> >
> >> I am looking forward to getting a reply from you.
> >>
> >>
> >> Brett
> >>
> >>
> >>
> >>
> >>
> >>
> >> At 2016-04-14 20:02:43, "Mark Abraham" 
> wrote:
> >> >Hi,
> >> >
> >> >If this "separation" was real, would you have expected to see
> >> >
> >> >a) a massive spike in the potential energy
> >> >b) complete inability for the constraint algorithm to enforce bond
> >> lengths?
> >> >
> >> >Hint, you haven't seen that.
> >> >
> >> >Mark
> >> >
> >> >On Thu, Apr 14, 2016 at 11:52 AM Justin Lemkul 
> wrote:
> >> >
> >> >>
> >> >>
> >> >> On 4/14/16 5:35 AM, Brett wrote:
> >> >> > Dear All,
> >> >> >
> >> >> >
> >> >> >
> >> >> > Thanks the reply.
> >> >> >
> >> >> >
> >> >> > But I think I need to discuss with you further on whether I need to
> >> >> discontinue my production MD. As I mentioned, resi 366-367 separated
> >> from
> >> >> the major body of the protein. By pymol I find resi 366-367 were at
> one
> >> >> edge of the box and were almost outside of the water box, and its
> >> >> neighbouring residues resi 365 and resi 368 were at the opposite
> edge of
> >> >> the box and were also almost outside of the water box.
> >> >> >
> >> >> >
> >> >> > It seems that way, although at the initial step by editconf I have
> put
> >> >> the protein in the center of the cubic box (1.5 nm), during the
> first 6
> >> ns
> >> >> the protein has moved significantly in the box, and at some moment
> resi
> >> >> 366-367 have moved out of the box, leading to the disconnection of
> resi
> >> >> 366-367 from the major body of the protein and 

Re: [gmx-users] further on: very strange phenomenon for my production MD by gromacs

[Brett]

Dear Mark,



My protein was a 6-identical subunit protein complex, initially each subunit 
contains a ligand. I want to see the conformational change from ligand bond to 
ligand free state, during this process there were reports that significant 
conformation change would occur


First from the initial PDB with the ligands, I delete the ligands and get the 
apo pdb, after pdb2gmx etc, I process the energy minimization, nvt 
equilibration and npt equilibration, then I start the production md. During the 
initial 6 ns production md, there were some conformation change observed, but I 
do not know whether it diffuse.


In addition, I do not understand what is "the envelope of the diffusion", and 
does diffusion leads the box from cubic to sphere?


Brett









At 2016-04-15 21:30:12, "Mark Abraham"  wrote:
>Hi,
>
>On Fri, Apr 15, 2016 at 3:18 PM Brett  wrote:
>
>> Dear All,
>>
>> By xmgrace I have checked the xmgrace, I find the spike was between
>> -8.245e+06 and -8.225e+06, is the spike massive or not?
>>
>>
>> In addition, by
>> gmx trjconv -f ma_0_1.trr -s md_0_1.tpr -o 6000pswithwater.pdb --pbc
>> nojump -dump 6000,
>>
>>
>> I got the 6 ns with water pdb and I have checked the pdb (with water box),
>> I find the resi 366-367 were not disconnected!
>>
>>
>> However, by gmx editconf, the original water box was in cubic, after "gmx
>> trjconv -f ma_0_1.trr -s md_0_1.tpr -o 6000pswithwater.pdb --pbc nojump
>> -dump 6000", the cubic water box was replaced by a rough sphere water box
>> with the protein roughly in the center. Will you please explain why the
>> cubic water box was replaced by the round sphere water box? Anything wrong
>> I have made?
>>
>
>You've simulated a bunch of molecules that diffuse. What shape would you
>expect for the envelope of the diffusion? How would it vary over time?
>
>Further, by visual check if in a production MD I did not see residue
>> disconnection caused by "Periodic_Boundary_Conditions", can we guarantee no
>> minor disconnection occurred during the md (or significant bond length
>> variance unobservable by the naked eye with pymol) if we did not process
>> the trjconv correction step?
>>
>
>The forces produced by such a disconnection would blow everything up
>immediately. Since it didn't...
>
>Mark
>
>
>> I am looking forward to getting a reply from you.
>>
>>
>> Brett
>>
>>
>>
>>
>>
>>
>> At 2016-04-14 20:02:43, "Mark Abraham"  wrote:
>> >Hi,
>> >
>> >If this "separation" was real, would you have expected to see
>> >
>> >a) a massive spike in the potential energy
>> >b) complete inability for the constraint algorithm to enforce bond
>> lengths?
>> >
>> >Hint, you haven't seen that.
>> >
>> >Mark
>> >
>> >On Thu, Apr 14, 2016 at 11:52 AM Justin Lemkul  wrote:
>> >
>> >>
>> >>
>> >> On 4/14/16 5:35 AM, Brett wrote:
>> >> > Dear All,
>> >> >
>> >> >
>> >> >
>> >> > Thanks the reply.
>> >> >
>> >> >
>> >> > But I think I need to discuss with you further on whether I need to
>> >> discontinue my production MD. As I mentioned, resi 366-367 separated
>> from
>> >> the major body of the protein. By pymol I find resi 366-367 were at one
>> >> edge of the box and were almost outside of the water box, and its
>> >> neighbouring residues resi 365 and resi 368 were at the opposite edge of
>> >> the box and were also almost outside of the water box.
>> >> >
>> >> >
>> >> > It seems that way, although at the initial step by editconf I have put
>> >> the protein in the center of the cubic box (1.5 nm), during the first 6
>> ns
>> >> the protein has moved significantly in the box, and at some moment resi
>> >> 366-367 have moved out of the box, leading to the disconnection of resi
>> >> 366-367 from the major body of the protein and finally leading to resi
>> >> 366-367 at the opposite of  its neighbouring residues resi 365 and resi
>> 368
>> >> in the cubix box.
>> >> >
>> >> >
>> >> > Am I right?
>> >> >
>> >>
>> >> Diffusion is normal.  The residues are not actually disconnected.  You
>> can
>> >> use
>> >> trjconv to "fix" this state, but there is nothing physically wrong with
>> it.
>> >> mdrun cares about physics, not our visualization convenience.
>> >>
>> >> -Justin
>> >>
>> >> >
>> >> > Brett
>> >> >
>> >> >
>> >> >
>> >> >
>> >> >
>> >> >
>> >> >
>> >> >
>> >> >
>> >> > At 2016-04-14 16:31:20, "Mark Abraham" 
>> wrote:
>> >> >> Hi,
>> >> >>
>> >> >> As the link Justin gave you says, and Tsjerk has since confirmed,
>> this
>> >> is
>> >> >> normal. There are infinitely many equivalent representations of your
>> >> >> simulation, not all of which will look connected for the thing that
>> is
>> >> of
>> >> >> interest.
>> >> >>
>> >> >> Strategies for handling visualisation are also on that link, but you
>> >> don't
>> >> >> need to handle anything within the simulation.
>> >> >>
>> >> >> Mark
>> >> >>
>> >> >> On Thu, 14 Apr 2016 09:48 Brett 

Re: [gmx-users] further on: very strange phenomenon for my production MD by gromacs

[Mark Abraham]

Hi,

On Fri, Apr 15, 2016 at 3:18 PM Brett  wrote:

> Dear All,
>
> By xmgrace I have checked the xmgrace, I find the spike was between
> -8.245e+06 and -8.225e+06, is the spike massive or not?
>
>
> In addition, by
> gmx trjconv -f ma_0_1.trr -s md_0_1.tpr -o 6000pswithwater.pdb --pbc
> nojump -dump 6000,
>
>
> I got the 6 ns with water pdb and I have checked the pdb (with water box),
> I find the resi 366-367 were not disconnected!
>
>
> However, by gmx editconf, the original water box was in cubic, after "gmx
> trjconv -f ma_0_1.trr -s md_0_1.tpr -o 6000pswithwater.pdb --pbc nojump
> -dump 6000", the cubic water box was replaced by a rough sphere water box
> with the protein roughly in the center. Will you please explain why the
> cubic water box was replaced by the round sphere water box? Anything wrong
> I have made?
>

You've simulated a bunch of molecules that diffuse. What shape would you
expect for the envelope of the diffusion? How would it vary over time?

Further, by visual check if in a production MD I did not see residue
> disconnection caused by "Periodic_Boundary_Conditions", can we guarantee no
> minor disconnection occurred during the md (or significant bond length
> variance unobservable by the naked eye with pymol) if we did not process
> the trjconv correction step?
>

The forces produced by such a disconnection would blow everything up
immediately. Since it didn't...

Mark


> I am looking forward to getting a reply from you.
>
>
> Brett
>
>
>
>
>
>
> At 2016-04-14 20:02:43, "Mark Abraham"  wrote:
> >Hi,
> >
> >If this "separation" was real, would you have expected to see
> >
> >a) a massive spike in the potential energy
> >b) complete inability for the constraint algorithm to enforce bond
> lengths?
> >
> >Hint, you haven't seen that.
> >
> >Mark
> >
> >On Thu, Apr 14, 2016 at 11:52 AM Justin Lemkul  wrote:
> >
> >>
> >>
> >> On 4/14/16 5:35 AM, Brett wrote:
> >> > Dear All,
> >> >
> >> >
> >> >
> >> > Thanks the reply.
> >> >
> >> >
> >> > But I think I need to discuss with you further on whether I need to
> >> discontinue my production MD. As I mentioned, resi 366-367 separated
> from
> >> the major body of the protein. By pymol I find resi 366-367 were at one
> >> edge of the box and were almost outside of the water box, and its
> >> neighbouring residues resi 365 and resi 368 were at the opposite edge of
> >> the box and were also almost outside of the water box.
> >> >
> >> >
> >> > It seems that way, although at the initial step by editconf I have put
> >> the protein in the center of the cubic box (1.5 nm), during the first 6
> ns
> >> the protein has moved significantly in the box, and at some moment resi
> >> 366-367 have moved out of the box, leading to the disconnection of resi
> >> 366-367 from the major body of the protein and finally leading to resi
> >> 366-367 at the opposite of  its neighbouring residues resi 365 and resi
> 368
> >> in the cubix box.
> >> >
> >> >
> >> > Am I right?
> >> >
> >>
> >> Diffusion is normal.  The residues are not actually disconnected.  You
> can
> >> use
> >> trjconv to "fix" this state, but there is nothing physically wrong with
> it.
> >> mdrun cares about physics, not our visualization convenience.
> >>
> >> -Justin
> >>
> >> >
> >> > Brett
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> > At 2016-04-14 16:31:20, "Mark Abraham" 
> wrote:
> >> >> Hi,
> >> >>
> >> >> As the link Justin gave you says, and Tsjerk has since confirmed,
> this
> >> is
> >> >> normal. There are infinitely many equivalent representations of your
> >> >> simulation, not all of which will look connected for the thing that
> is
> >> of
> >> >> interest.
> >> >>
> >> >> Strategies for handling visualisation are also on that link, but you
> >> don't
> >> >> need to handle anything within the simulation.
> >> >>
> >> >> Mark
> >> >>
> >> >> On Thu, 14 Apr 2016 09:48 Brett  wrote:
> >> >>
> >> >>> Dear All,
> >> >>>
> >> >>> As I have introduced in my previous e-mail,
> >> >>>
> >> >>> "After energy minimization and equilibrations, I am now running a
> 50 ns
> >> >>> production MD, for a protein of 6 identical subunits, with each
> subunit
> >> >>> about 300 residues (from resi 120 to resi 420), and there were no
> >> breaks in
> >> >>> any chain. Every day it runs about 1 ns, and every day I use the
> >> command
> >> >>> trjconv to get a new PDB based on the md_0_1.trr file, for the
> >> comparison
> >> >>> between the new pdb and the initial pdb.
> >> >>>
> >> >>>
> >> >>> Today I got the PDB at 6 ns. However when I checked it my pymol, I
> find
> >> >>> there is something very strange. Although the rmsd between the 6 ns
> md
> >> PDB
> >> >>> and 0ns md PDB was about only 3.7, for chain B, residue 366-367
> moved
> >> 180
> >> >>> angstrom away from this residues neighbouring residues, and
> >> correspondingly
> >> >>> make chain B 

[gmx-users] further on: very strange phenomenon for my production MD by gromacs

[Brett]

Dear All,

By xmgrace I have checked the xmgrace, I find the spike was between -8.245e+06 
and -8.225e+06, is the spike massive or not?


In addition, by
gmx trjconv -f ma_0_1.trr -s md_0_1.tpr -o 6000pswithwater.pdb --pbc nojump 
-dump 6000,


I got the 6 ns with water pdb and I have checked the pdb (with water box), I 
find the resi 366-367 were not disconnected!


However, by gmx editconf, the original water box was in cubic, after "gmx 
trjconv -f ma_0_1.trr -s md_0_1.tpr -o 6000pswithwater.pdb --pbc nojump -dump 
6000", the cubic water box was replaced by a rough sphere water box with the 
protein roughly in the center. Will you please explain why the cubic water box 
was replaced by the round sphere water box? Anything wrong I have made?


Further, by visual check if in a production MD I did not see residue 
disconnection caused by "Periodic_Boundary_Conditions", can we guarantee no 
minor disconnection occurred during the md (or significant bond length variance 
unobservable by the naked eye with pymol) if we did not process the trjconv 
correction step?


I am looking forward to getting a reply from you.


Brett






At 2016-04-14 20:02:43, "Mark Abraham"  wrote:
>Hi,
>
>If this "separation" was real, would you have expected to see
>
>a) a massive spike in the potential energy
>b) complete inability for the constraint algorithm to enforce bond lengths?
>
>Hint, you haven't seen that.
>
>Mark
>
>On Thu, Apr 14, 2016 at 11:52 AM Justin Lemkul  wrote:
>
>>
>>
>> On 4/14/16 5:35 AM, Brett wrote:
>> > Dear All,
>> >
>> >
>> >
>> > Thanks the reply.
>> >
>> >
>> > But I think I need to discuss with you further on whether I need to
>> discontinue my production MD. As I mentioned, resi 366-367 separated from
>> the major body of the protein. By pymol I find resi 366-367 were at one
>> edge of the box and were almost outside of the water box, and its
>> neighbouring residues resi 365 and resi 368 were at the opposite edge of
>> the box and were also almost outside of the water box.
>> >
>> >
>> > It seems that way, although at the initial step by editconf I have put
>> the protein in the center of the cubic box (1.5 nm), during the first 6 ns
>> the protein has moved significantly in the box, and at some moment resi
>> 366-367 have moved out of the box, leading to the disconnection of resi
>> 366-367 from the major body of the protein and finally leading to resi
>> 366-367 at the opposite of  its neighbouring residues resi 365 and resi 368
>> in the cubix box.
>> >
>> >
>> > Am I right?
>> >
>>
>> Diffusion is normal.  The residues are not actually disconnected.  You can
>> use
>> trjconv to "fix" this state, but there is nothing physically wrong with it.
>> mdrun cares about physics, not our visualization convenience.
>>
>> -Justin
>>
>> >
>> > Brett
>> >
>> >
>> >
>> >
>> >
>> >
>> >
>> >
>> >
>> > At 2016-04-14 16:31:20, "Mark Abraham"  wrote:
>> >> Hi,
>> >>
>> >> As the link Justin gave you says, and Tsjerk has since confirmed, this
>> is
>> >> normal. There are infinitely many equivalent representations of your
>> >> simulation, not all of which will look connected for the thing that is
>> of
>> >> interest.
>> >>
>> >> Strategies for handling visualisation are also on that link, but you
>> don't
>> >> need to handle anything within the simulation.
>> >>
>> >> Mark
>> >>
>> >> On Thu, 14 Apr 2016 09:48 Brett  wrote:
>> >>
>> >>> Dear All,
>> >>>
>> >>> As I have introduced in my previous e-mail,
>> >>>
>> >>> "After energy minimization and equilibrations, I am now running a 50 ns
>> >>> production MD, for a protein of 6 identical subunits, with each subunit
>> >>> about 300 residues (from resi 120 to resi 420), and there were no
>> breaks in
>> >>> any chain. Every day it runs about 1 ns, and every day I use the
>> command
>> >>> trjconv to get a new PDB based on the md_0_1.trr file, for the
>> comparison
>> >>> between the new pdb and the initial pdb.
>> >>>
>> >>>
>> >>> Today I got the PDB at 6 ns. However when I checked it my pymol, I find
>> >>> there is something very strange. Although the rmsd between the 6 ns md
>> PDB
>> >>> and 0ns md PDB was about only 3.7, for chain B, residue 366-367 moved
>> 180
>> >>> angstrom away from this residues neighbouring residues, and
>> correspondingly
>> >>> make chain B has a break at residue 366-367!"
>> >>>
>> >>> A moment ago by trjconv I regot the6 ns pdb with water, I find the
>> residue
>> >>> 366-367 are almost (or exactly) at the edge of the water box (but by
>> naked
>> >>> eye it seems the residue 366-367 are not outside of the water box). In
>> this
>> >>> does the moving away of the residue 366-367 are still caused by "
>> >>> Periodic_Boundary_Conditions", and can I continue the production md to
>> >>> completion with the residue 366-367 disconnected from the major part
>> of the
>> >>> protein complex and then I correct it by trjconv?
>> >>>
>> >>> I am