I failed in changing the job.sh-File, so at the moment I try to make a simulation only with Em_1.mdp/NVT_1.mdp/NPT_1.mdp to learn. Everything seemed to work fine until this:
gmx_mpi mdrun -s NPT.tpr -deffnm NPT
I got no xvg and cpt output-file. There was only a memory data error. I don’t know, if my description is enough to help and maybe I made bigger mistakes in using the tutroial on other data. But I have no idea right now and would appreciate help very much.
If you’re getting LINCS warnings, the solution is not to disable constraints (which for GROMOS are required for all bonds). Those warnings indicate instability in the system that should be diagnosed further.
Note that the tutorial inputs are incorrect for GROMOS anyway, as the nonbonded cutoffs are different (the tutorial is for OPLS-AA).
Thank you very much for your answer! Now I used:
constraints = all-bonds
rlist=rcoulomb=rvdw= 1.4
rvdw-switch = 1.3
I used a cubic simulation box of 2.8 nm.
I’m so sorry, but I need to ask further questions. When I start job.sh my output is only Lambda_0, but there’s no production-run gro-file. It seems, as if only Lambda_0 is done again and again without a complete production-run. End of production run Lambda_0 and beginning of new Lambda_0 looks like that:
GROMACS: gmx mdrun, version 2018
Executable: …/gromacs/gromacs_bin/bin/gmx_mpi
Data prefix: …/gromacs/gromacs_bin
Working dir: …/GROMOS96_54a7/Lambda_0/Production_MD
Command line:
gmx_mpi mdrun -deffnm md0 -ntomp 64
Compiled SIMD: AVX2_256, but for this host/run AVX_512 might be better (see
log).
Reading file md0.tpr, VERSION 2018 (single precision)
Changing nstlist from 20 to 100, rlist from 1.4 to 1.4
No option -multi
Using 1 MPI process
Using 64 OpenMP threads
NOTE: The number of threads is not equal to the number of (logical) cores
and the -pin option is set to auto: will not pin threads to cores.
This can lead to significant performance degradation.
Consider using -pin on (and -pinoffset in case you run multiple jobs).
starting mdrun ‘linaloolwater’
500000 steps, 1000.0 ps.
Free energy home directory set to …/GROMOS96_54a7
.mdp files are stored in …/GROMOS96_54a7/mdp
Starting minimization for lambda = 0…
:-) GROMACS - gmx grompp, 2018 (-:
It’s not complaining Lambda_0 does already exist or something like that. When I delete NPT and the production-run in the Shell Script, only doing EM and NVT, it does all Lambdas. When I do EM, NVT and NPT it seems to have a problem in Lambda2, but I had no output for this because of connection problems. I just have no idea, where the problem could be.
My warnings are:
when insert-molecules: WARNING: all CONECT records are ignored
every time grompp: *WARNING 1 [file Em_0.mdp]: You are using full electrostatics treatment PME for a system without charges. This costs a lot of performance for just processing zeros, consider using Cut-off instead.
every mdrun Em, one LINCS warning: Step 9, time 0.009 (ps) LINCS WARNING
relative constraint deviation after LINCS:
rms 0.003932, max 0.057541 (between atoms 1647 and 1648)
bonds that rotated more than 30 degrees:
atom 1 atom 2 angle previous, current, constraint length
761 762 32.2 0.0971 0.1003 0.0972
762 763 32.0 0.0971 0.1004 0.0972
1097 1098 38.2 0.0970 0.1026 0.0972
1098 1099 40.6 0.0968 0.1022 0.0972
1646 1647 41.9 0.0969 0.1019 0.0972
1647 1648 37.7 0.0971 0.1028 0.0972
writing lowest energy coordinates.
Does a normal (non-free energy) simulation run stably? You’re having stability issues, meaning that either the topology is insufficiently accurate, the run settings are inappropriate, you have bad contacts, etc.
The warnings are all normal output; CONECT records are never used for anything. The topology is the definitive source of atomic linkage.
That was quite the best idea, you could give me. It doesn’t work! I just didn’t think of this. I still try to find the problem, but now it’s much easier to search for it, because I have more experience with normal simulations.
I thank you so much for always answering so fast and giving good advice!