Warren,
You're right that one can not use constraints if they are also using lincs on
C-C bonds.
However, constraining this bond is not necessary with a 1 fs timestep.
Even while constraining C-C bond lengths, one can still use harmonic restraints
instead (also from the pull code).
Before
Thanks David. I think that we're saying the same thing, but there is
indeed a 'correct' denatured state at a given temperature and pressure
and protein concentration and salt concentration, etc. It was this to
which I was referring.
Chris.
On 2010-10-18 06.56, chris.neale at utoronto.ca
Chen,
to answer this, please see David's post and my reply. To answer if
your unfolded state ensemble is accurate, you will need to know some
experimentally evaluated quantities of the state that you are trying
to reproduce: 498K, 26 atm, a certain amount of salt, etc, (but not
including
Generally, forcefields are not parameterized for temperatures other
than 298K, so simulations are not expected to reproduce the expected
properties (like boiling water and the correct temperature
denaturation of proteins).
There's almost certainly other issues here (including the fact that
Stefane,
for more information on what Mark is talking about here, check out his
publication:
M. J. Abraham and J. E. Gready (2008) Ensuring Mixing Efficiency of
Replica-Exchange Molecular Dynamics Simulations J. Chem. Theory Comput.
(4), 1119-1128.
Dear all,
I come back to you for several
Has anybody checked g_hbond results for consistency between v4.0.7 and v4.5.1?
My usage is:
g_hbond -f my.xtc -r 0.68 -contact -life -s my.tpr -ac
and I find that the output files hbnum.xvg, hbac.xvg, hblife.xvg, and
the ac-reported lifetime of interactions are all different. To show
the
If I redo this, selecting only a single acceptor atom and 64 donor
atoms, with a large cutoff distance of 3 nm, v4.0.7 gives me 44
should have read 4 nm above. the point is that there really are 44
contacts, but v4.5.1 reports only 1.
contacts in one frame while v4.5.1 gives me only 1
Stefane,
why are you hesitant to reduce the size of your system. If you can
still address the same questions with a smaller system, then I'd say
the answer is nearly always to use that smaller system (REMD or not).
Besides, my gut feeling is that even if you had enough 271*16 cpus
Thank you David, Berk, and Roland.
unix cat is a good option (I didn't realize that this was possible)
and so is scripting to do it in a number of steps.
I have filed enhancement request #593.
Thank you,
Chris.
--- original message ---
On 2010-10-13 01.54, chris.neale at utoronto.ca
Sanku,
This is not possible using g_energy or any other gromacs tool that I am aware
of.
You must calculate the temperature derivative of the free energy.
See, for example, equation 16 in Pettitt BM, Rossky PJ. J Chem Phys.
1986;84:5836
-- original message --
Hi,
I am studying Potential of
Stefane,
If you are asking for a comparison of simulation time required, then
that is not something that can be answered in the abstract. There are
even cases in which REMD will take more total simulation time to
converge the 300K ensemble than will N simulations all at 300K. The
efficiency
Sanku,
decomposing free energies into enthalpic and entropic components is not
my particular expertise, but I'm astounded to hear that enthalpy equals
the average forcefield energy. I would have thought that, at the very
least, pressure work would also need to be incorporated.
Honestly, I
Thank you for the insight Mark.
To get it on the record, this is fairly annoying for trjcat (not
trjconv). If anyone else has run into this, then please do post
because a memory efficient cat mechanism will only be coded if there
is sufficient interest.
But again, thanks for your good
Dear Zhongjin:
1. You have indeed provided more data, but please try to make some
progress on your own. E.g. you tell me the order of selections into
g_dist, but you should try reversing the order.
2. Your figures are difficult to interpret -- I only opened Snap5, but
I don't know what
Dear Zhongjin:
First, I see no reason why you would not expect your g_dist values to
be negative. If you want the positive distances, then reverse the
order of the groups as you pass then to g_dist. Also, what type of
values do you get directly out of mdrun and into the pull_x.xvg file?
I don't think that this is currently causing anybody any problems, but
note that genbox is going to cut any lipids that cross out of the
central unit box (either because genbox is unaware of PBC or because
these lipids now clash across PBC).
Therefore:
genbox -cs popc128a.pdb -o
Landry,
I replied earlier, but it didn't seem to make it to the list. If 2
show up then they are identical.
The mailing list archive post that you cite is actually talking about
a (3-point) angle and not a (4-point) dihedral angle.
If you want dihedral angles, then take a look at how
Landry,
you can't do it using the pull code, but it is possible. We did US
over Phi and Psi in the following paper, and you can get the methods
from there. I also posted many more specifics and examples to the
mailing list during that time.
C. Neale, T. Rodinger, and R. Pomès,
Aswathy:
We can't tell if you did US correctly unless you post what you have
done. You must give more info and you must copy and paste the actual
input / .mdp parameters that you used.
As for if your PMF is correct, then we can never tell you that. You
can, however, ensure that your
Manoj,
to suggest that Justin is being lazy would be to ignore the hundreds,
possibly thousands, of helpful posts that he has made on the gromacs
mailing list to assist users like yourself free of charge.
In fact, Justin was probably helping you more than you have realized
by pointing
Maria,
I highly doubt that you are correct. If you want more help, please
copy and paste your commands and some relevant output back to the list.
Chris.
--original message --
Message: 4
Date: Wed, 8 Sep 2010 15:07:15 +0200
From: maria goranovic mariagorano...@gmail.com
Subject: Re:
Maria, try this. There actually is a lot of this on the mailing list,
so I suggest checking it a little deeper for your next querry, or at
least outlining how you looked and what you found so that it is clear
you have tried.
Also, read about pull_pbcatomN and think carefully about how you
Sorry Alex, but there is no good substitute what I have already
suggested. I helped you find a lead and I suggested a way to simplify
your system to discover what is going on. Now it's up to you to do
that work.
It might end up being a bug, but I doubt it. I have done things
similar to
It's a messy solution, but you could
1. run trjconv so that your cylinder is in the unit cell
2. convert your trajectory to .gro files
3. massively expand the x and y dimensions of each .gro
4. join the .gros to a .xtc
5. g_rdf
You could also use template.c to write a program that does this
Can you post a snip of your pullx and pullf files from the water box
simulation where you pulled on 2 waters?
Quoting alexander yakovenko yakovenk...@ukr.net:
Yep, my case is b :(... Gromacs do writes pullx as ref COM global
coords and pull COM displacement. And I still don't realize what
Alex,
If you do decide to continue along with the method that I suggested,
I'll join back in. To this point, I see no indication that you have
tried to relate the force to the position for the water runs, which is
my standing suggestion to you.
Chris.
Quoting alexander yakovenko
Alexander:
Can you please also post your free energy profile out of wham? Also,
can you confirm that you are using different force constants for
different windows?
Justin is correct, but that just means that you are using more cpus
than you need... it's not actually wrong. What are your
Alexander, please post the following:
1. force constants for your plot at
http://picasaweb.google.ca/117558205101329348732/G_wham#5511287294315721970
NOTE: if they are all 1000 as in your first post, then you have a
problem in your sampling.
2. units for your x-axis in the plot mentioned
First, I don't think that you converted and debiased correctly. Once
you have a histogram of the displacement, then you will debias that
histogram like this (assuming run at 298K):
FC=1000
T=298
cat biased.hist |awk '{print
$1,-0.00831447*'${T}'*log($2)-0.5*'${FC}'*$1*$1}' debiased.hist
First, you confuse me when you say pulling direction was along
Y-axis since your .mdp file indicates pull_dim = Y Y Y and so you are
pulling in *all* dimensions.
Your profile doesn't look like mine at all, but then again I took mine
from a force profile. Now that I can see your forces and
So now we have the first real lead. Next is for you to simplify your
system to the point where you *can* understand the relationship
between the pullf and pullx files. For example, I always suggest a
water bath of explicit water with the sd integrator and apply the pull
code to two waters.
Dear Alex:
be careful with g_density! If you do not use the symmetrize -symm
option, then the box height is going to be taken as a single frame. If
you do NpT, then you will get an incorrect (or at least noisy) result.
I forget if this will bite you if you do use the -symm option.
This
Gavin, I recall mentioning gen_vel=no with temperature coupling as a
possible problem a while ago (the problem is that the initial forces
become initial velocities and then those get scaled up and what you
have are velocities and not temperatures -- read about the flying ice
cube problem).
Dear Nimesh:
trjconv -f orig.xtc -s my.tpr -o new.xtc -pbc atom
to get the help information:
trjconv -h
Normally, you might not get an answer to this one since it's the most
commonly asked question. The users list search is, however, not
working right now so I thought I'd pass along the
I think that there is some problem with the users list search on the
gromacs page: http://www.gromacs.org/Support/Mailing_Lists/Search
A gromacs users list search for molecule returns zero results.
A search for gromacs does return lots of results.
A google search for gromacs and molecule
velocities are not taken from the .gro as far as I know. You can get
them from the .trr if you also use a .edr.
If you want to check, then do a run for only 1 ps and save your .edr
every timestep. Then run g_energy and look at the temperature.
I am not sure that a constant energy
Natalie:
It would also be a good idea for you to ensure that your supervisor
knows how difficult it is to learn and correctly apply these methods.
If they think that a person can learn linux and gromacs and the
gromacs pull code in a short period of time, then the stage is set for
I have seen this before, and there are a few possible reasons, but I'm
still waiting to see that histogram with two peaks
(http://lists.gromacs.org/pipermail/gmx-users/2010-August/053311.html). It's
really hard to help you when we have to guess what the problem looks
like... or perhaps you
Yup, just got the link. Thank you for posting, and sorry that I missed
it. It seems possible that Justin is correct and you'll get good data
if you take it out of the pullx.xvg file (perhaps the g_dist tool is
wrapping pbc for you).
1. How big is your box definition in your .gro file?
You
Dear Gavin:
What happens if you replace your reference group by a single water
molecule and your pulled group by a single water molecule and run the
2.35 nm simulation again. Do you get the same two-peaked histogram?
What about if you use a system with a single water for the reference
Tsjerk,
Two reasons. First, I simply don't understand awk as well as you do.
Second, I think that breaking it down into simpler commands linked by
pipes is better for teaching. If somebody needs help to combine 3 cols
into a vector distance, then they may copy and paste your solution to
Gavin,
thank you for plotting the sampling by the time. Looks like this isn't
a two state problem. Can you also show a histogram image on-list
please? Just a histogram of the sampled distances.
Thanks,
Chris.
Sorry forgot to attach the files
Hi Chris
Sorry for the delay. Please find
The .edr file contains the instantaneous energies at every nstenergy steps.
-- original message --
Hello,
Does gromacs write the current values (at some requested step) to
ENER.EDR or the averaged ones (at the requested interval)?
--
gmx-users mailing listgmx-users@gromacs.org
As I said before, please plot the position vs. time. It would be best
if you can make a graph of this and post it to some web tool. I use
photobucket, but there are many others.
I realize that you think you are starting at equilibrium, but there
are always things that can have unexpected
Gavin, I wish you the best of luck. Perhaps somebody else on the list
can help you.
-- original message --
Thanks Chris
I have equilibrated the system at 600 K in a previous run therefore
would it not be equilibrated, and also would the velocities from the
final configuration of the
Your language was not rude, but I did get frustrated after twice
asking you to perform an analysis and twice being ignored. My advice
still stands: plot the deviation vs. time and report back to the list.
Chris.
-- original message --
Cheers and thanks for the suggestions.
P.S. I hope you
Dear Gavin:
plot the position vs. time. Probably you will see it sample around one
value, then discover a new state, and make a sharp transition to
another value. If the new state is much more favourable than the
initial state, then you can discard the initial state as
equilibration. It
Dear Xueming:
the word mol is short form for mole
http://en.wikipedia.org/wiki/Mole_%28unit%29
In the pull code context, it refers to moles of the pulled group.
The force is not applied to the COM of a cluster. The magnitude of
the force is determined based on the COM distance, and then the
This surprised me since most of the systems that I am interested in
have improper dihedrals. Is it possible that I did something wrong here?
### Error message from mdrun-gpu:
Program mdrun-gpu, VERSION 4.0.99-dev-20100315
Source code file:
Thank you Rossen, that worked.
When I compile, should I be including my FFTW as per normal? I didn't
do that since it was not mentioned and during the build, I get --
Using internal FFT library - fftpack
--- original message ---
Hi Chris,
## what I did:
module purge
module load
This sounds like you, or your sysadmin, killed the job. It is also
possible if you are using a cluster that there is a walltime limit and
you exceeded that, so your job was sent the term signal.
You should restart the run. Check the manual on how to do that since
it depends on what version
This surprised me since most of the systems that I am interested in
have improper dihedrals. Is it possible that I did something wrong here?
### Error message from mdrun-gpu:
Program mdrun-gpu, VERSION 4.0.99-dev-20100315
Source code file:
I haven't tried mdrun -nt based hyperthreading, but I have tested
using -np 16 on an 8 core box. I get an 8% to 18% performance increase
when using -np 16 and optimizing -npme as compared to -np 8 and
optimizing -npme. This is on a cluster of Intel Xeon E5540 aka
Nehalem with 2 quad cores
Sounds like you killed the job, or your sysadmin killed it, or it
timed out if you are on a cluster with a walltime limit.
-- original message --
Hi all,
I have been running a simulation of 200 ns length and it suddenly
stoped at 67 ns showing the following line -
Received the TERM
Nisha,
The approach is dictated by the goal. What do you want from this? and
why are you doing it? e.g. if you want to test the FF, then it is a
good idea to at least include US as a part of your strategy; if you
want to determine if 100 ns of equilibrium sampling is sufficient to
Nisha:
Simply applying available tools and seeing if the distribution looks
okay is not a good plan. You should have a very well-defined idea of
what you are trying to test and then pick a tool to get that done.
For example, perhaps you have a reference that provides the converged
Dear Szilárd and Rossen:
I have attempted with gcc4.4.0 and I got the same result. I can verify
that cmale/FindCUDA.cmake exists and contains the MARCO for
CUDA_INCLUDE_DIRECTORIES but it seems that this library is not, for
some reason, getting loaded in by at least
Dear users:
I previously posted about difficulties compiling gromacs 4.5b2 for GPUs:
http://lists.gromacs.org/pipermail/gmx-users/2010-August/052905.html
It turns out that this problem was not directly related to version.
First, I had to add -DLIB_INSTALL_DIR=$(pwd)/install_dir to the cmake
Thanks Rossen, I did not yet get it working. I just posted under a
more relevant title, please see the post:
http://lists.gromacs.org/pipermail/gmx-users/2010-August/052939.html
(there were 2 errors, one that I fixed and one that I am still
struggling with).
Chris.
-- original message --
It turns out that I can make the error message with
CUDA_INCLUDE_DIRECTORIES go away by adding the following text to
kernel/gmx_gpu_utils/CMakeLists.txt
MACRO(CUDA_INCLUDE_DIRECTORIES)
FOREACH(dir ${ARGN})
SET(CUDA_NVCC_INCLUDE_ARGS ${CUDA_NVCC_INCLUDE_ARGS} -I${dir})
ENDFOREACH(dir
Dear all:
I have an explicit solvent peptide-in-bilayer system taht also used
the pull-code.
This system runs fine on 8 cores in a single box with v4.0.7 and
4.5b2. It also runs fine on 16 cores IB for v4.0.7 but not 4.5b2. In
the later case, I get an exit from mdrun (see below). I have
It seems that cmake 2.8.0 is not recent enough?
I attempted to compile like this:
cd /project/pomes/cneale/GPC/exe/intel/gromacs-4.5-beta2
module purge
module load intel
module load cuda #cuda 3.0
module load cmake#cmake 2.8.0
export CC=icc
export CXX=icpc
cmake
Ah, I see that I need version =2.6.4
This info is on the web, but not in the INSTALL_GPU file... perhaps
that file could include all of the information found here:
http://www.gromacs.org/index.php?title=Download_%26_Installation/Related_Software/Gromacs_on_GPUs
Thanks,
Chris.
-- original
Dear Aswathy:
When you post to the list, you should include a description of what
you eventually want to do. Here's an example of why this is important.
If you want to make PMF then you must use the values of the pulled
group. I'm not promoting using SMD to make PMFs, but since it is
Dear Alan:
I don't have a solution for you, but I do have a simpler test case: a
single spc water molecule. I have not addressed your noted energy
difference between programs because I am not familiar with them
(although you should try gromacs compiled in double precision).
I can
I'd send a list of links except that the other end of a google search for
gromacs pull code example
has all of what you will need to broaden your general pull code
understanding. The mailing list has an absolutely massive number of
posts on this too.
Chris.
Hi all,
Could anybody please
Susan. Please use the mailing list instead of personal email.
A google search for electronegativity of selenium provides a useful
first hit. The .rtp issue is covered often on the mailing list.
Overall, you will find this link helpful:
Dear Ehud:
A quick google search turned up:
http://www.bmrb.wisc.edu/ref_info/atom_nom.tbl
# A.A. BMRBSC PDB UCSFMSI XPLOR SYBYL* MIDAS*
DIANA
# - - -
-
G HA2 pro-R
Dear Shuai:
When you get some advice, it's useful if you can try it before coming
back to the list.
Vitaly was simply giving you an idea about how to proceed. You need an
initial state that it high-density and this is one way to get such a
state fairly easily: start with a low density
Depending on your system and chosen thermostat, gen-vel=no can be a
dangerous idea (very small original velocities simply get scaled up);
besides, you should be able to equilibrate from a gen-vel=yes pretty
quickly. I don't disagree with Justin's comment, but just because it
works doesn't
Dear Jagannath:
There is, as far as I know, no way to fix this. Here's what you should do.
1. Let the space between your timesteps be X ps.
2. Use trjconv -pbc cluster -dump X -o out.gro
2b. if that doesn't work, try trjconv -pbc cluster -dump (X*2) -o out.gro
2c. if that doesn't work, try
PS:
you realize that trjconv -pbc cluster is never going to work unless
you actually have a cluster, right? So you can't start this procedure
near your initial dispersed state. You need to start at some time T
where you actually do have a cluster. Then you can try -dump T, -dump
(T+X),
Dear shuai shuai:
Please use a descriptive title. I renamed it.
You must describe your problem more. e.g. is the problem that you get
a frozen cluster in some simulations, or that you don't get this in
other simulations? What, exactly, is a small cluster of frozen
molecules?
You must
Dear Shuntaro:
If you're using PME then there is no need to have unit charge in
charge groups. I can't comment on any other errors that may or may not
be in that topology as I have never simulated ATP. Also, if you're not
using PME, then I can not advise you if it is ok to have non-integer
That depends. mdrun pull code does it smartly via pull_pbcatom
(assuming that you have chosen pull_pbcatom and your box size
smartly), while g_dist is not smart about it at all. Can you be more
specific about what you want to do?
-- original message --
Hi everyone:
How gromacs calculate
I wrote my own program to do this and so I never tried that one. We
did find that clustering based on a dingle atom in the chain was a
very good idea. So take your distal carbon in the surfactant and make
in index group of it and cluster it (with a larger cutoff) with
g_clustsize.
The pullx file should contain everything that you need. dX,dY,dZ *are*
the components of the distances between the two groups.
I still don't understand your question (perhaps you could give us a
worked example with real data?), but if you just want the COM
distance, then try this:
cat
perhaps that's not standard terminology, sorry. The numbering starts
at the headgroup so the distal atom would be the one at the end of the
acyl-chain (the one most likely to spend its time in the middle of the
cluster.
-- original message --
Thanks. Just wondering what is a 'distal atom'
Gavin, let's take g_wham out of the picture for now to simplify
things. First, you umbrella is pretty strange. You want them to have a
distance of 0.0 nm (pull_init1 = 0)???
So let's look at making a histogram from your raw data.
1. cat pullx.xvg |awk '{print sqrt($1,$5*$5+$6*$6+$7*$7)}'
Gavin,
I think it's time for you to read some papers and read the gromacs
manual pull COM section. A few days of reading will make all of this
clear to you.
-- original message --
Hi Chris
Thanks very much for your help. The reason I used pull_init1 = 0 is
because this is the value that
It seems that I do not need to care about decimal total group charges
as I am planning to use PME.
-- well, find it in the literature first!
-- original message --
Dear Chris,
I really appreciate your comment.
Can you provide a reference for the idea that the charge groups must
have
I missed that pull_start = yes, sorry. So Gavin, your run should have
stayed around the starting position. Make the histogram directly and
you should get a better idea.
Chris.
-- original message --
To clarify, the tutorial sets pull_init1 = 0 because it is used in
conjunction
with
I don't have the parameters. You could possibly fudge it by
extrapolating out from SER-CYS-Se_CYS based scaled by
electronegativities of O,S,Se. I'm not saying that this is rigorous,
just that its possible to do. My main point here, though, is to ask if
you *really* need to bother with
Shay: that's a better idea than I had, thanks.
Tsjerk: that's likely as fast as its ever going to get. this is great.
I was initially worried that it would mangle the last line of the .gro in a triclinic box definition, but it seems to be ok.
Thanks all,
Chris.
-- original message --
What
If you have a directory full of .gro files that contain velocities,
and if you don't need those velocities, then you're taking up too much
space on your drive.
It's too bad, but trjconv -novel -f in.gro -o out.gro will still write
the velocities to out.gro.
Here's a small script to strip
Hi guys,
I just wanted to chime in on the idea that how you perform the
initial pulling to get structures isn't all that important because I
strongly disagree and think that it is, in general, immensely important.
OK, it's actually not important if we're talking about pulling two Na+
Sounds correct to me. One can use any method to generate the starting
conformations. Some methods may yield equilibrium starting structures
faster than other methods, but they are all ok in principle.
Nevertheless, as with any simulations, one must ensure that care is
taken to maintain
Dear Gavin:
1. Please use an informative title like pull_group0 moves
unexpectedly so that those searching the list later can find it useful.
2. We're using Newton's equations, so equal and opposite forces are
what you should expect. Why is there a problem is pull_group0 moves?
Chris.
In addition to what everyone else has said, please note that pull_dim
= N N Y is going to allow X and Y distance components to get as
large/small as they will without any restraint. This may be what you
want, but if you want a distance, but want it to be mostly in Z, then
pull_dim = Y Y Y
multiple timesteps are not possible as far as gromacs 4.0.7. NAMD can do this.
-- original message --
Is it possible to carry out multiple time step molecular
dynamics simulations
in Gromacs
4.0. versions ? Could you
please give me some information about this issue ?
Thank you very much for
Can you please supply gen_temp and also your pressure coupling values?
Also, please supply the g_energy values that show the temperatures
changing as you suggest (with accompanying timestep info).
I'm fairly sure that what is happening is that you are simply seeing
the nose-hoover
I recall seeing something online about how gromacs developers have
decided to focus on increasing the overall speed and allowing
generally large timesteps (via e.g. angle constraints) vs.
implementing multiple timestepping (no mailing list ref. sorry). I
agree that this is not a logically
Carla: Agreed that this can be confusing.
What you seem to want can be obtained with g_anaeig -filt
Nevertheless, What you get with -proj can also be very useful.
For a very quick overview: -filt will output a .xtc file that contains
only those motions that are along the selected EVs.
Dear Elio:
I think I was pretty clear in my last email:
If you still have trouble, then your next post should include things
like I tried this... but I didn't understand this specific part
to make it clear that you have put in the effort here.
Otherwise, hopefully somebody else has the
Dear Elio:
First, you need to convert your PMF to a free energy difference
between states. That means defining a boundary between, for example,
bound and unbound states. The integrate your PMF over the bound and
unbound states to get two dG values, and their difference is your free
There's lots of information in the gromacs manual to show that the
.atp file is not the only file that you must modify. My most important
suggestion to you is that you should read the entire manual, possibly
twice, before attempting to build new molecules.
Gromacs manual, page 96:
5.3.1
Dear Thanasis:
8 posts on the same topic? I didn't even read them. If you want some
help, I'd suggest that you start with a new subject line and post a
single email containing all relevant information.
You can see how it looks at the bottom of this list:
Please clarify:
Are doing SMD or are you doing US? If you're doing SMD then you should
not be using WHAM and you should not really be able to generate any
sampling histograms.
Are the histograms that you are referring to population densities of
the sampling along your reaction
You're right, although I did not realize this at first. I got confused
(again) by the global numbering. Regarding the off-by-one numbering, I
can figure out what is meant there and that was not really my question.
Thanks Berk, and sorry for the confusion.
Chris.
I coded this and I looked
Take that replica on the left that shows a bimodal histogram. Now plot
a time series of the displacement: x-axis = time and y-axis =
displacement along reaction coord. Is it jumping back and forth
between two regions of sampling? Probably not... more likely it starts
near one maximum and
201 - 300 of 816 matches
Mail list logo