Could not read agb file.

[vijay singh]

I have the params.agb file in the "gromacs/share/gromacs/top". The md.log still gives the error -

Could not read agb file. GBSA is being omitted.

I am running with CUDA, 64 bit linux with PR4.

Any suggestion on what I could do?

Thanks,

[Peter Eastman]

Is it possible you have Gromacs installed in two different places, and you've put the file in the wrong one? I've seen that happen before when people have several versions of Gromacs.

Peter

[vijay singh]

Thanks Peter for the reply. There were a few different versions of gromacs installed on our system. I just removed the redundant ones that I had access to.

It is not very clear, from the error message on the log file, which directory it is seeking in -

Could not find/read params.agb file needed for OBC/GBSA -- aborting: lib directory=< y>.

When I do "echo $libdir", I get a blank. Using print statements, it appears that this output is being issued from "/src/kernel/md_openmm.cpp" - lines 277 - 289.

I will also make a request with the computing cluster staff to see if we can have just one gromacs install.

Thanks for your time,
Vj

[Mark Friedrichs]

Try setting the environment variable GMXLIB to the full pathname of the directory containing the agb file

[vijay singh]

I did have the GMXLIB set to the full pathname. I also added the full pathname to PATH. Not sure if that would really matter. But now the nature of the problem seems to have changed to "Segmentation fault" when I use the mdrun_openmm.

[Romica Sandu]

Hello!

I happened to have - "Segmentation fault" when I use the mdrun_openmm -
when the original structure was insufficiently minimized. You can convince this trying the simulation with mdrun

There were and other reasons but
I forgot. However, eventually you find one solution experiencing.

[vijay singh]

Thanks for the tip. I should have updated earlier, the Segmentation error is not showing up now. When I run "mdrun_openmm" using "gdb debugger", the commands go through fine except for a line in the md.log -

agb parameter file line=<> is being skipped

From an earlier post (https://simtk.org/forum/message.php?msg_id=2194) I gather this can be ignored.

If I don't use "gdb", I am not getting any output. There seems to be some power issues here, I have not been able to login to the system for last few hours. I believe my problem might be with some settings and I should double check to make sure.

Thanks...

[Peter Eastman]

Try running it in gdb to find out where the segmentation fault is occurring.

You may get a few "illegal instruction" errors in places with names like nb_kernel_ia32_3dnow_test_asm(). Those are a normal part of how Gromacs starts up, so just hit "c" to continue. Once it stops at the segmentation fault, type "bt" to get a stack trace.

Peter

[Silvio a Beccara]

Vijay, can I ask you which force field you are using with the Generalized Born model?

Silvio

[Mark Friedrichs]

Hi Silvio,

The OpenMM GB model is based on the paper listed below. The OpenMM implementation was validated against Tinker's implementation of OBC. The overlap scale factors used in the program were obtained from Tinker (ksolv.f). The atomic radii used in the algorithm are listed in the 'params.agb' file. The first column in the file is the atom type, and the second is the radius in Angstroms. The other columns are ignored. We have been using the amber94 and amber99 force fields (different radii for the two force fields).

A. Onufriev, D. Bashford and D. A. Case, "Exploring Protein Native States and Large-Scale Conformational Changes with a Modified Generalized Born Model", PROTEINS, 55, 383-394 (2004)