I'm waiting on email conformation for the GMX user group, but this is in reply to Reviewers so I'm pressed for time and asking here.
I'm building a small collagen-based system. The crystal structure dimensions are:
15.91920 2.58660 2.70113 0.00000 0.00000 0.00000 0.00000 -0.18557 0.00000
I'm using Gromacs version 2016.4 - I can't change this as it's what's installed on the work server.
I've built this system before (with slight changes) and I've never had these problems. I keep getting segmentation faults when running gmx mdrun. There is no output in the .log file beyond the system configuration output. I do, however, see the Tolerance (Fmax) and Number of steps before mdrun dies.
My execution line:
gmx mdrun -deffnm enrun -rdd 1 -nt 1 -pin on
Non-verbose screen output:
Back Off! I just backed up enrun.log to ./#enrun.log.3#
Running on 1 node with total 24 cores, 24 logical cores
Hardware detected:
CPU info:
Vendor: Intel
Brand: Intel(R) Xeon(R) Gold 6126 CPU @ 2.60GHz
SIMD instructions most likely to fit this hardware: AVX_512
SIMD instructions selected at GROMACS compile time: SSE2
Hardware topology: Only logical processor count
Compiled SIMD instructions: SSE2, GROMACS could use AVX_512 on this machine, which is better.
Reading file enrun.tpr, VERSION 2016.4 (single precision)
Using 1 MPI thread
Using 1 OpenMP thread
At the moment my geometry file does not contain water. That's fine - I was just testing my geometry at this stage. However, when I try solvating the system with:
gmx solvate -cp test.gro -cs spc216.gro -p system.top -o solvated2.gro
I get another segmentation fault:
Reading solute configuration
CRYSTAL STRUCTURE OF COLLAGEN-LIKE PEPTIDE AT 1.27 ANGSTROM
Containing 7822 atoms in 640 residues
Reading solvent configuration
216H2O,WATJP01,SPC216,SPC-MODEL,300K,BOX(M)=1.86206NM,WFVG,MAR. 1984
Containing 648 atoms in 216 residues
.....
Generating solvent configuration
Will generate new solvent configuration of 9x2x2 boxes
Solvent box contains 13827 atoms in 4609 residues
Removed 3036 solvent atoms due to solvent-solvent overlap
Segmentation fault
The fact that gmx solvate crashes is a lot more revealing than gmx mdrun. I'm just a little unsure of how to troubleshoot these issues?
Much appreciated. Thanks.
Did you run all the regression tests? It seems to me that there's something fundamentally wrong with the GROMACS installation (and certainly suboptimal, given the note about SIMD in your output). Have you compiled your own version from source and tested it? Often pre-built binaries are problematic and HPC centers often do not keep up to date or do proper testing.
Thanks Justin for taking the time to reply.
Your suggestion of installing my own distribution on the server is going to be the best solution.
I had the strangest similar bug, that I could solve by just changing a little the order in which bonds/angles/dihedrals appear. I could reproduce my bug by putting back the initial order of the bonds/angles/dihedrals I was defining in the ITP file. Basically there was no reason at all for this bug, in terms of system setting. That was using 2018.4 gromacs version.
Right! I've remotely connected to my home machine to try and distribution that I know works and I've tried one other distribution at work.
2019.2 (mine) - fail
2016.4 - fail
2018.2 - fail
BUT...
I made an empty .gro file, 10x10x10, and gmx solvate works on each of these distributions.
So, given that gmx solvate does not consult FF data (at a guess), it must come purely down to the .gro file content?
Something to note, I have a post-translation modified amino acid in the structure. I added this by hand modifying the forcefield data and it goes straight through pdb2gmx with no problems. Taking that into account, as a control I have successfully solvated the non-PTM structure (the wildtype i.e., the structure before I made the post-translational modification to the Lysine-Argine side chains).
My next questions/paths or inquiry would be whether customised structures affect gmx solvate? I've done this before with other PTM structure. Clearly I've messed this one up!
EDIT:
It's the content of the .gro file. Something is causing gmx to die. I took out everything apart from a single NPRO. I adjusted the number of atoms and ran gmx solvate. It solvated successfully.
Then I only took out the PTM amino acid. It goes back to failing. I'm not sure what in a .gro file would cause gmx solvate to crash.
It is also worth noting, trying to load the structure which I'm trying to solvate into VMD causes VMD to crash. However, avogadro happily loads it.
I think I know what the problem is.
I used a final structure for an earlier stimulation. There are 27 polypeptide chains not connected so they split all over the box (see attached). That's what I've used as my starting structure for pdb2gmx and all other gmx operations. I've preserved the box dimensions, so in theory, I shouldn't have a problem (I think). When I remove parts outside the box dimensions, gmx solvate works.
I've managed to restore the structure using pbc using:
trjconv -pbc atom
The structure is attached. Sadly gmx solvate is still failing.
The gro files have NaN for the hydrogens built by pdb2gmx for the PTM residue (search for NGLY in the gro files). That probably comes from the heavy atoms for that residue having 0 for all Y and Z coordinates. Things should be ok once that's fixed!
Dear Mark Abraham I have a question about REMD simulation with implicit solvent
1- when I want to run equilibration I have this error that when you use sn-type = simple, DD does not support simple you should use grid or use one thread.
I change the command to :
gmx mdrun -nt 1 -deffnm nvt
I confront with this error (segmentation fault)- core dumped.
I upload my .mdp file could you please help me what should I do? I use the 2018.8 version of gromacs.
2-when I want to run REMD simulation with implicit solvent-based on your tutorial.
I run the equilibration section but when I see the result such as ener.edr file the value of the temperature is (nan) in all directories. I do not know what is the problem.
could you please help me what should I do? I really need to do this REMD simulation with an implicit solvent.
please take a look at to my mdp file which I upload here.
Many thanks
Hi, please ask these questions on gmx-users mailing list. I don't work for GROMACS any more and those that do are there
- Badges
- Science topic