Dear Ania Di Pede Mattatelli

My understanding is that you produced a new binary (step07a_npt.tpr) from a new .mdp file (step07a_npt.mdp), and now you are trying to continue the previous nvt run (step06d_nvt.*0, right?

In that case this is wrong

$ gmx_mpi mdrun -v -s bin/tpr/step07a_npt.tpr

-cpi bin/cpt/step06d_nvt_ann_LONG.cpt

-o bin/trj/step06d_nvt_ann_LONG.trr

-x bin/trj/step06d_nvt_ann_LONG.xtc

-cpo bin/cpt/step07a_npt.cpt -cpt 60

-c str/complex/step07a_npt.gro

-e bin/edr/step06d_nvt_ann_LONG.edr

-g run/log/mdlog/step06_nvt_ann_LONG.log

You are using a NEW mdp file, and generated a NEW binary. The info from the previous state should be already inside the new binary, that's why in the preprocessor phase you wrote the flags -c, -r, and -t. Therefore this

$ gmx_mpi mdrun -v -s bin/tpr/step07a_npt.tpr

should do the work. Consider that -o, -x etc specify the OUTPUT options, not the input, therefore when you write

-x bin/trj/step06d_nvt_ann_LONG.xtc

you are telling mdrun to write your trajectory to that file, not to take from that file and continue it. The error you get in the first case is related to the fact that the -cpi option is a continuation from a checkpoint, which is going to search for some files to continue a given run. This is useful when a simulation crashes or when you want to make a simulation longer. For example, you run 100ns of something and would like to go to 150ns. Then you run gmx conver-tpr, you extend the tpr file time and continue from there (using -cpi flag, read https://manual.gromacs.org/documentation/current/user-guide/managing-simulations.html)

Here you got a new tpr, so you don't have to specify the -cpi flag.

Second error is similar. You are going from NVT to NpT, therefore you are now coupling the pressure. What mdrun is telling you is that it is not able to go on with the NVT phase because the tpr you are using (the step07) has something different from the state you are trying to continue (step06), most likely the barostat but something else as well for sure.

If you want to make the NVT longer you can go

gmx convert-tpr -s npt.tpr -extend TIME_YOU_WANT_TO_EXTEND_IN_PS

and then go

gmx mdrun -v -s npt.tpr -cpi npt.cpt

In this case you want a NpT, so a change in the binary, therefore you (rightly) compiled a new tpr (specifying the input status with -c/-r/-t flags) and can now just run the

gmx mdrun -v -s npt.tpr

Hope this is clear! :)

Nicola

Hi, Nicola Piasentin ! Thank you very much for your help!!!

I tried your suggestion and it worked! :D

But I had to do it all on the main folder (at the bottom I wrote my new workflow).

All the outputs went in the same folder.

Seems that now gmx doesn't complain about changing the simulation algorithm in the mdp file either (as I needed to go from heating_nvt ---> equilibrium_npt ---> production MD).

Is there a possibility to keep files outputs going into different subdirectories?

I am really uncomfortable with all gromacs files in the same folder.

I am trying to divide files into subfolders organized by format.

bin/ctp, bin/edr, bin/tpr, bin/trj, top/itp, top/top, str/complex/xxx.gro, and so on...

________________________________________________________________________________

---> WORKING PROCEDURE:

> gmx_mpi grompp -f par/mdp/step07a_npt.mdp

-c str/complex/step06d_nvt_ann_LONG.gro

-r str/complex/step06d_nvt_ann_LONG.gro

-p top/top/test.top

-t bin/cpt/step06d_nvt_ann_LONG.cpt

-n str/idx/step06c_complex.ndx

-o bin/tpr/step07a_npt.tpr

-pp top/top/step07a_npt_POSRE.top

-po run/log/mdout/step07a_npt_mdout.mdp

I had to cp the tpr from the bin/tpr folder to the main folder and run next steps from there (can't make it work from the subdirectories).

---> equilibrium npt md run:

> gmx_mpi mdrun -v -deffnm step07a_npt -cpt 60

output in the main folder:

step07a_npt.cpt step07a_npt.edr step07a_npt.gro step07a_npt.xtc

---> MD production starts:

> gmx_mpi grompp -f par/mdp/step08a_md1.mdp

-c step07a_npt.gro -r step07a_npt.gro

-n str/idx/step06c_complex.ndx

-p top/top/test.top

-t step07a_npt.cpt

-po run/log/mdout/step08a_md1_mdout.mdp

-pp top/top/step08a_md1.top

-o step08a_md1.tpr -maxwarn 100

> gmx_mpi mdrun -v -deffnm step08a_md1 -cpt 60 -nsteps 10

output in the main folder:

step08a_md1.cpt step08a_md1.edr step08a_md1.gro step08a_md1.xtc

---> I did only 10 steps as a test and then continued on a cluster submission.

> job.slurm:

... [options] ...

mpirun -np 96 gmx_mpi mdrun -v -s step08a_md1.tpr -cpi step08a_md1.cpt -deffnm step08b_md1 -cpt 60

Dear Ania Di Pede Mattatelli

Glad I helped!

About how to split the files, that's up to you. I think this depends on running with the flag -deffnm, which is telling gromacs to do all with that name. This means gromacs will look for the file NAME.tpr and produce output according to NAME.cpt, NAME.edr, NAME.gro etc.

Probably you can do what you want by dropping the -deffnm file and specify the output for everything, such as

gmx mdrun -v -s /path/to/tpr -o /path/for/output/files -cpo /path/to/checkpointfile

You can see the options by running gmx mdrun --h, under the line 'Options to specify output files'. I don't think you will be able to send the .edr to an .edr directory, .log to .log directory etc, at least by looking at the available flags.

Personally I don't split files based on their file types but rather on run types. I prefer to keep a general directory where I have a directory for the forcefield (.itp/.prm/ etc), one for topology and indexes, and then those for the runs, something like

./1_em/

./2_NVT/

./3_NpT/

./4_prod/

./ff/

./top_ndx/

However, this is up to you, but I feel like you will have only more confusion if anything by splitting per file type. Consider for example you want to run a quick analysis on the NVT run. You cd inside, and there you have your .edr, .xtc, .tpr, etc. which will give you fast access to everything you need (.tpr in particular). Most of analysis requires several flags to specify structure/trajectory/binary/whatever, therefore each time you have everything inside the same directory where you cd'ed. Think also about compiling new prod run

cd ./4_prod/ (where you have prod.mdp or similar)

gmx grompp

-f ./prod.mdp -o prod.tpr

Thanks Nicola Piasentin for the suggestion!

I am a beginner user of gmx and MD. I just wanted to keep my experiment clear. Your way is far more straitforward, I will borrow it! Thanks!