Path Collective Variable

[Anurag Sethi]

Hi,

I am using PCVs to study a large conformational change in a large
protein (>400 residues). I ran a targeted molecular dynamics and was
able to get the conformational change to occur. Then I used PCVs along
41 conformations along the path. These conformations were chosen such
that the distance between neighboring conformations was nearly the
same along the whole path. Then I calculated lambda to be the
expectation value of mean sq. deviation between neighbors. Then I ran
the simulation with about 41 replicas (at each of these positions).

After close to 0.7 ns per replica, I observe that some of the replicas
are able to move in "s" but some are stuck in the same "s" value. For
example, replica 1 varies between s value of 1 and 2 while replica 2
is stuck at an s value of 2. I also post processed the TMD run to find
that the 's' and 'z' along the path in the TMD simulation and found
that the pathway has a z value between 2 and 3 through most of the
path except at one spot close to s=25 where z goes upto 6. How can I
improve the conformations I choose along the path?

Secondly, when is RMSD based PCV a good idea versus contact distance
based or DRMSD based PCV?

Thanks,
Anurag

[Davide Branduardi]

Hi!

> I am using PCVs to study a large conformational change in a large
> protein (>400 residues). I ran a targeted molecular dynamics and was
> able to get the conformational change to occur. Then I used PCVs along
> 41 conformations along the path. These conformations were chosen such
> that the distance between neighboring conformations was nearly the
> same along the whole path. Then I calculated lambda to be the
> expectation value of mean sq. deviation between neighbors.

Which code are you using? Are you using MSD / DMSD /CMAP ?
In case of MSD: remember that for gromacs the lambda should be in nm^-2 so if you calculate rmsd with VMD those values should be adjusted
accordingly

> Then I ran
> the simulation with about 41 replicas (at each of these positions).

Normal MD, metadynamics or steered?

>
> After close to 0.7 ns per replica, I observe that some of the replicas
> are able to move in "s" but some are stuck in the same "s" value.

If you are doing normal MD might be due to a very low diffusion coefficient along that CV. There is not much you can do about it: you need to simulate more.
If you are doing a metadynamics probably it must be due to a exceedingly large lambda (this is true when your value stick to "integer" values like 2.000 3.00 4.000 etc...)

> For
> example, replica 1 varies between s value of 1 and 2 while replica 2
> is stuck at an s value of 2. I also post processed the TMD run to find
> that the 's' and 'z' along the path in the TMD simulation and found
> that the pathway has a z value between 2 and 3 through most of the
> path except at one spot close to s=25 where z goes upto 6.

In what units? Ang? nm? What CV?
Can you first add these details since it is not really clear to me what you are doing and why.
Ciao
D

> How can I
> improve the conformations I choose along the path?
>
> Secondly, when is RMSD based PCV a good idea versus contact distance
> based or DRMSD based PCV?
>
> Thanks,
> Anurag
>

> --
> You received this message because you are subscribed to the Google Groups "PLUMED users" group.
> To post to this group, send email to plumed...@googlegroups.com.
> To unsubscribe from this group, send email to plumed-users...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/plumed-users?hl=en.
>

*********************************************
Davide Branduardi
Postdoc
Theoretical Molecular Biophysics Group
Office 2.255
Max Planck Institute for Biophysics
Max-von-Laue-Straße 3
60438 Frankfurt am Main
Phone: +49 069 63031503
Fax: +49 069 63031502
davide.b...@biophys.mpg.de
davide.b...@gmail.com
Web:
https://sites.google.com/site/davidebranduardi
*********************************************

[Anurag Sethi]

Davide,

Thanks for your reply and sorry for not putting in the details.


I am using MSD at the moment. So, when I calculated the average MSD
between neighbors and equated lambda to 2.303/<MSD between neighbors>,
lambda = 3.9 Angstrom^-2.

When I ran the simulation with the 41 replicas, it was a metadynamics
run with the PCVs as the collective variables. So, when I was running
the metadynamics run with these 41 replicas, some replicas were stuck
at the same value of s but others were moving a little more freely
along the "s" direction. I am running it further. I just wanted to
know if this is normal behavior. The sigma for the gaussians added
along the path (PCV s) was 0.2 based on the average standard deviation
of "s" during 10 different normal MD simulations with very different
conformations along the pathway.

While I figuring out how to put the upper boundary on z, I went back
to the original TMD run to figure out how much the TMD path deviated
from the path I defined in terms of the 41 frames. That is, I measured
s and z for all frames in the TMD run. During the TMD run, the system
periodically visited regions with z = 2-3 Angstrom^2 in between the
chosen frames (where z is nearly 0). There was one region in between
frames 25 and 26 when the system visited z=6. I just wanted to know if
this was normal too.

In general, I wanted a better idea of when to use MSD, DMSD, and CMAP
while using PCV to explore the free energy landscape (metadynamics or
umbrella sampling) during conformational changes in proteins.

Thanks,
Anurag


On Mar 12, 4:23 am, Davide Branduardi <davide.brandua...@gmail.com>
wrote:

> > For more options, visit this group athttp://groups.google.com/group/plumed-users?hl=en.

>
> *********************************************
> Davide Branduardi
> Postdoc
> Theoretical Molecular Biophysics Group
> Office 2.255
> Max Planck Institute for Biophysics
> Max-von-Laue-Straße 3
> 60438 Frankfurt am Main
> Phone: +49 069 63031503
> Fax:      +49 069 63031502

> davide.brandua...@biophys.mpg.de
> davide.brandua...@gmail.com
> Web:https://sites.google.com/site/davidebranduardi
> *********************************************

[Davide Branduardi]

Hi Anurag,

On Mar 12, 2012, at 2:56 PM, Anurag Sethi wrote:

> Davide,
>
> Thanks for your reply and sorry for not putting in the details.
>
>
> I am using MSD at the moment. So, when I calculated the average MSD
> between neighbors and equated lambda to 2.303/<MSD between neighbors>,
> lambda = 3.9 Angstrom^-2.
>

So I assume you're playing with NAMD (or some other angstrom-based code)

> When I ran the simulation with the 41 replicas, it was a metadynamics
> run with the PCVs as the collective variables. So, when I was running
> the metadynamics run with these 41 replicas, some replicas were stuck
> at the same value of s but others were moving a little more freely
> along the "s" direction.

What you mean by stuck? In my previous mail I asked you if by "stuck" you meant
at discrete values (like 4.00 5.00 6.00 ) and not somewhere in between (like 6.345678).
If they are stuck at discrete values then you have a problem that your lambda is too high.
Another point: is your sigma for metad well chosen? try to figure out if they fit with the expected fluctuation
you obtain out of a free dynamics plot.
Nevertheless, if the system is large and you have lots of atoms in the path, your free energy landscape
might appear very rough and therefore you should expect several configurations to be stuck somewhere.

> I am running it further. I just wanted to
> know if this is normal behavior. The sigma for the gaussians added
> along the path (PCV s) was 0.2 based on the average standard deviation
> of "s" during 10 different normal MD simulations with very different
> conformations along the pathway.

Ok, so the problem with sigma is not to be taken into account.

>
> While I figuring out how to put the upper boundary on z, I went back
> to the original TMD run to figure out how much the TMD path deviated
> from the path I defined in terms of the 41 frames. That is, I measured
> s and z for all frames in the TMD run. During the TMD run, the system
> periodically visited regions with z = 2-3 Angstrom^2 in between the
> chosen frames (where z is nearly 0).

It is good but this is already telling me that probably your lambda is too high.
Take this alternative version for the choice of lambda
lambda=2.3/largest msd between two adjacent frames
does is change significantly from the previous choice?
Generally I make so to have low Z values covered by metadynamics between two adjacent frames.

> There was one region in between
> frames 25 and 26 when the system visited z=6. I just wanted to know if
> this was normal too.

That is a bit too large. But consider that this must be an artifact of your initially chosen targeted dynamics.
In fact targeted paths tends to cut corners and be far from ideal paths. So my choice would be to make your reaction
occurr through a smd along the path and rechoose frames from that trajectory in an iteratively fashion
(ADVERT Moment: check what we did for http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0032397 )
That is generally improving a lot but after a while you cannot improve further (so you have to slow down your steering speed
or move with a free energy calculation)

>
> In general, I wanted a better idea of when to use MSD, DMSD, and CMAP
> while using PCV to explore the free energy landscape (metadynamics or
> umbrella sampling) during conformational changes in proteins.

it is strongly dependent on your problem.
My opinion is that if I want to induce some specific rotation of bonds and simultaneously bond formation I go with MSD.. If it is about putting
sticky patches together then CMAP is good. dMSD is somewhat intermediate. But probably the CMAP and DMSD expert can give an additional
word on it.
Ciao CIao

D

> For more options, visit this group at http://groups.google.com/group/plumed-users?hl=en.
>

*********************************************
Davide Branduardi
Postdoc
Theoretical Molecular Biophysics Group
Office 2.255
Max Planck Institute for Biophysics
Max-von-Laue-Straße 3
60438 Frankfurt am Main
Phone: +49 069 63031503
Fax: +49 069 63031502

[qingh...@gmail.com]

Hello plumeders,

I have some questions about the path collective variables, so I just post on a very old post.
I apologize if I am not doing the right way.

Should all frames in the reference pdb file be aligned? For the starting structure of the simulations, is it necessary that it is one frame of the reference structures (key conformations are the same)? And is it necessary to do multiple replicas of simulations, like Anurag did? Thanks very much!


All the best,
Qinghua

[Giovanni Bussi]

Hi,

1. It is not necessary to pre-align frames. Running frames are aligned at every step.

2. If I understand correctly from the old email replicas were used for multiple walkers. I would say it is not necessary, though it might allow you to increase statistics.

Giovanni

To unsubscribe from this group and stop receiving emails from it, send an email to plumed-users+unsubscribe@googlegroups.com.

To post to this group, send email to plumed...@googlegroups.com.

Visit this group at https://groups.google.com/group/plumed-users.
To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/d8a15351-1811-4870-81ee-9c6885ed04f9%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[qingh...@gmail.com]

Dear Prof. Bussi,

Thanks very much for your comments.

I tried to run a simulation with path collective variable, but it failed. Here is the input file for plumed:

# Unit setting
UNITS LENGTH=A TIME=ps ENERGY=kcal/mol
# make chain AB whole
    chainAB: GROUP ATOMS=1-8258
    WHOLEMOLECULES ENTITY0=chainAB
#
path: PATHMSD REFERENCE=refs.pdb LAMBDA=25.0
#
PRINT STRIDE=10 ARG=* FILE=colar.dat

What I got in the colar.dat file is something like:

#! FIELDS time path.sss path.zzz
 0.000000 -nan inf
 0.040000 -nan inf
 0.080000 -nan inf
 0.120000 -nan inf

The reference structures are from different windows of a umbrella sampling, and the RMSD looks fine

The reference file I prepared is also attached. Could you please have a check for me? Thanks very much!


All the best,
Qinghua

To unsubscribe from this group and stop receiving emails from it, send an email to plumed-users...@googlegroups.com.

[Massimiliano Bonomi]

Looks like your pdb file is formatted in a strange way.
If I am not mistaken, atoms name should occupy column from 14 to 16.
Max

> On Apr 5, 2017, at 16:41, qingh...@gmail.com wrote:
>
>
>
> Dear Prof. Bussi,
>
> Thanks very much for your comments.
>
> I tried to run a simulation with path collective variable, but it failed. Here is the input file for plumed:
>
> # Unit setting
> UNITS LENGTH=A TIME=ps ENERGY=kcal/mol
> # make chain AB whole
> chainAB: GROUP ATOMS=1-8258
> WHOLEMOLECULES ENTITY0=chainAB
> #
> path: PATHMSD REFERENCE=refs.pdb LAMBDA=25.0
> #
> PRINT STRIDE=10 ARG=* FILE=colar.dat
>
> What I got in the colar.dat file is something like:
>
> #! FIELDS time path.sss path.zzz
> 0.000000 -nan inf
> 0.040000 -nan inf
> 0.080000 -nan inf
> 0.120000 -nan inf
>
> The reference structures are from different windows of a umbrella sampling, and the RMSD looks fine
>
>
>
>

> To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/97478daf-5a18-40e4-ab01-c646d8d46fcb%40googlegroups.com.

> For more options, visit https://groups.google.com/d/optout.

> <refs.pdb>

[Giovanni Bussi]

It looks like the file you sent have dos terminations.

Can you try with the one attached and let me know it it fixes the problem?

Giovanni

To unsubscribe from this group and stop receiving emails from it, send an email to plumed-users+unsubscribe@googlegroups.com.

To post to this group, send email to plumed...@googlegroups.com.
Visit this group at https://groups.google.com/group/plumed-users.

To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/97478daf-5a18-40e4-ab01-c646d8d46fcb%40googlegroups.com.

[Massimiliano Bonomi]

Sorry, you are right, it’s column 13 to 16.
Got distracted by the different alignment of some of the atoms…

Max

[qingh...@gmail.com]

Dear Prof. Bussi,

Thanks very much for your fast respond. Yeah, it is working now with your file.
It is strange that I got the dos things even I edited the file under linux.

About the path collective variable, I have another question.
As I want to explore the pathway of a loop opening and closing, I am wondering
whether I should also explore some ranges beyond the crystal open and close states?
What do you think of it? Thanks very much!


All the best,
Qinghua

[qingh...@gmail.com]

Hello Max,

Thanks very much for your reply! All the pdb files in the reference file were generated by pdb2gmx, then awk, and vim.
So I guess that the pdb file should be in the correct pdb format. Thanks to Prof. Bussi, he helped me fix the problem.


All the best,
Qinghua

[Giovanni Bussi]

The simplest way is to add an upper wall on zzz so as to keep the trajectory close to the pathway.

The downside of this is is that you will not discover alternative paths. Additionally, if you constraint too much the path you might affect the entropic barriers.

In any case, what you are trying to do is not trivial, check the literature for similar applications!

Final question: with which version of PLUMED did you have the dos problem? I though this was fixed.

Giovanni

To unsubscribe from this group and stop receiving emails from it, send an email to plumed-users+unsubscribe@googlegroups.com.

To post to this group, send email to plumed...@googlegroups.com.
Visit this group at https://groups.google.com/group/plumed-users.

To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/a15b8998-909e-4f6d-9e89-11157c62e244%40googlegroups.com.

[qingh...@gmail.com]

Dear Prof. Bussi,

Thanks very much for your suggestions. I will read some paper with similar situations, and run some test simulations to see how it is going.

To be consistent with previous simulations, I am performing the simulations with Gromacs-4.6.7 and plumed-2.1-hrex.

I already have the plumed 2.3 installed, I will use the new one for other projects.

All the best,
Qinghua