- Posts: 17
initram-v5.sh seg fault
- edesantis
- Topic Author
- Offline
- Fresh Boarder
I am trying to convert the Martini representation of a protein crystal into a gromacs one.
in the first minimization steps, I have a segmentation fault error.
reading in the grompp log, I found this warning
WARNING: Listed nonbonded interaction between particles 238941 and 239092
at distance 3.034 which is larger than the table limit 3.000 nm.
This is likely either a 1,4 interaction, or a listed interaction inside
a smaller molecule you are decoupling during a free energy calculation.
Since interactions at distances beyond the table cannot be computed,
they are skipped until they are inside the table limit again. You will
only see this message once, even if it occurs for several interactions.
IMPORTANT: This should not happen in a stable simulation, so there is
probably something wrong with your system. Only change the table-extension
distance in the mdp file if you are really sure that is the reason.
since this is my first attempt to doing the backmapping, I don't know how to proceed,
any of you have ideas?
thanks in advance
Emiliano
Please Log in or Create an account to join the conversation.
- peterkroon
- Offline
- Gold Boarder
- Posts: 210
Gromacs is not a forcefield or model, so trying to "convert the Martini representation of a protein crystal into a gromacs one" is nonsensical. Beyond that, could you give us the exact commands you used?
Peter
Please Log in or Create an account to join the conversation.
- edesantis
- Topic Author
- Offline
- Fresh Boarder
- Posts: 17
sorry, I meant gromos not gromacs
the command I've used is:
./initram-v5.sh -f CG_file.gro -o aa_file.gro -p aa_topology.top -to gromos
the conversion from GC to aa is done (projected.gro file was created), but after about a hundred of steps of sd minimisation I have this segmentation fault without any other message written,
thank you again
Emiliano
Please Log in or Create an account to join the conversation.
- tsjerk
- Offline
- Expert Boarder
- Posts: 103
Did you make the molecules whole before backmapping?
Cheers
Please Log in or Create an account to join the conversation.
- edesantis
- Topic Author
- Offline
- Fresh Boarder
- Posts: 17
gmx trjconv –pbc no jump
gmx trjconv -pbc mol
should I use also -whole?
thanks
Emiliano
Please Log in or Create an account to join the conversation.
- edesantis
- Topic Author
- Offline
- Fresh Boarder
- Posts: 17
but the seg fault problem is still present...
Emiliano
Please Log in or Create an account to join the conversation.
- riccardo
- Offline
- Platinum Boarder
Did you take a look (e.g., with VMD, pymol) at the "projected.gro" - whether it looks reasonable or not (e.g., part of the projected protein is far away from the rest of it or things like this)?
Please Log in or Create an account to join the conversation.
- edesantis
- Topic Author
- Offline
- Fresh Boarder
- Posts: 17
the box is quite big (29nm *35nm* 15nm) and it is very crowded (there are 500 lyzosyme), so looking at the details it is not so easy,
but overall the projection seem to be done in the rigth way,
if a superimpose with VMD the projection and the CG representations there is good overlapping among them
Emiliano
Please Log in or Create an account to join the conversation.
- jbarnoud
- Offline
- Admin
- Posts: 23
If it is the second one, then you probably need to run the first one longer with the '-em' option. If the first minimization crashes, increasing the initial random kick with '-kick' may help. In some rare cases, running on a single CPU can help; use the '-mp' option to do so.
Please Log in or Create an account to join the conversation.
- edesantis
- Topic Author
- Offline
- Fresh Boarder
- Posts: 17
it is the first one that is crashing,
I've already tried to run the minimisation using a single CPU by typing the command invoked by initram directly from the terminal but even in that case the minimisation crashed,
I will try now changing the kick value
bye
Emiliano
Please Log in or Create an account to join the conversation.
- edesantis
- Topic Author
- Offline
- Fresh Boarder
- Posts: 17
I've tried to increase the kick value (in several steps, increasing each time of one order of magnitude) and the segmentation fault error is still appearing.
So I decided to cut off a portion of the crystal from my simulation box, to exclude that the problem could arise from the dimension of my system.
Even using only one unit cell (5292 beads), I have, after some minimisation steps of the first minimisation run the seg fault.
I've also tried to run the simulation with the previous version of gromacs (4.5.5). In this case, both the minimisations ended because the steepest descent converged to the machine precision but did not reach the requested Fmax (so it seems there is not any problem)
But, as I start with the md simulation, I have a lot of LINCS warning (even if I use a lincs order =16) and the simulations crashed.
Looking at the gro and trr generated during the two minimisation, I can see that most angles and bonds are adjusted, but there still are some weird superimposition of atoms, that are the cause of the lincs warnings
do you have any other suggestions to overcome that problem??
are there other tools to perform the backmapping??
thanks a lot,
Emiliano
Please Log in or Create an account to join the conversation.
- jbarnoud
- Offline
- Admin
- Posts: 23
If the minimization works with gromacs 4.5.5, then there may be things to check to make the simulations themselves work. The most obvious thing to try would be to minimize furter with the non-bonded interactions turned on. If I well remember, you can increase the number of steps of that second minimization with the "-nb" option. As your system is large, you can go to very high numbers in the order of thousands or even tens of thousand of steps.
I guess produced the protein topology using martinize. If it is the case, it should have portions that are only parsed by grompp if the "FLEXIBLE" flag is set. If you produced the topology in any other way, or if you modified it, make sure that you have these portions. Constraints do not minimize well, so the FLEXIBLE flag replaces them by stiff bonds.
Please Log in or Create an account to join the conversation.
- edesantis
- Topic Author
- Offline
- Fresh Boarder
- Posts: 17
to have more minimization step is possible,
but the will not change since both minimisations ended prematurely for having reached the machine precision
the protein topology for the cg representation was produced using the martinize.py, while the topology I give to initram-v5.sh to do the backmapping (-p flag) was produced using pdb2gmx starting from the same pdb I've used as input for martinize.py
Is it possible to attach the gro file in the forum?
(I didn't find the way to do that...)
thank you for your help
Emiliano
Please Log in or Create an account to join the conversation.
- jbarnoud
- Offline
- Admin
- Posts: 23
I read your previous message to fast and did not notice why the minimization stopped. Indeed, asking for more steps would be useless. And I mixed up what topology was used when, it is indeed not the CG one that is relevant here.
Maybe you can try to run the simulation with a shorter time step, 0.1 fs instead of 0.2. This is ridiculously slow, but there are major frustrations to solve and there are a lot of them as your system is large.
You can also run the simulation by saving every integration step. Visual inspection of that trajectory may give you clues on what causes the crash.
By the way, at that step of the backmapping, it is not worrisome to have some LINCS warning assuming it does not stop the simulation. Indeed, the first equilibration steps are there to solve these issues.
Please Log in or Create an account to join the conversation.
- tsjerk
- Offline
- Expert Boarder
- Posts: 103
Can you send me a link to the coarse grained structure, all required atomistic topology stuff and mapping files if you made your own?
Cheers,
Tsjerk
Please Log in or Create an account to join the conversation.
- edesantis
- Topic Author
- Offline
- Fresh Boarder
- Posts: 17
desycloud.desy.de/index.php/s/00l7nejqipPkQ4D
at this link there is all the stuff required for the conversion
thanks,
Emiliano
Please Log in or Create an account to join the conversation.
- edesantis
- Topic Author
- Offline
- Fresh Boarder
- Posts: 17
jbarnoud wrote: It is odd that you get a crash on gromacs 2016 but not 4.5 and this requires some investigation on our side. If you have a gro file that you could share with us so we can reproduce the error it would help us fixing the issue.
If the minimization works with gromacs 4.5.5, then there may be things to check to make the simulations themselves work. The most obvious thing to try would be to minimize furter with the non-bonded interactions turned on. If I well remember, you can increase the number of steps of that second minimization with the "-nb" option. As your system is large, you can go to very high numbers in the order of thousands or even tens of thousand of steps.
I guess produced the protein topology using martinize. If it is the case, it should have portions that are only parsed by grompp if the "FLEXIBLE" flag is set. If you produced the topology in any other way, or if you modified it, make sure that you have these portions. Constraints do not minimize well, so the FLEXIBLE flag replaces them by stiff bonds.
Hi,
did you have the opportunity to have the investigations from your side?
thank in advance,
Emiliano
Please Log in or Create an account to join the conversation.
- jbarnoud
- Offline
- Admin
- Posts: 23
I am sorry you case got buried in my TODO list. I had a look, and there seems to be some issues with your initial CG structure. In the structure with one unit cell, the proteins have steric clashes with the next periodic image. This , for sure, is going to cause issues.
There is a lot going on with your larger system, so it is difficult to say if you have the same problem there; but you probably do.
Are your CG system the output of a simulation? If so you should have a look at how you deal with PBC. If not, your box may be too small to accommodate your system.
I hope it helps,
Jonathan
Please Log in or Create an account to join the conversation.