Answers to frequently asked questions (FAQs)¶

Questions regarding GROMACS installation¶

Do I need to compile all utilities with MPI?

With one rarely-used exception (pme_error), only mdrun is able to use the MPI parallelism. So you only need to use the -DGMX_MPI=on flag when configuring for a build intended to run the main simulation engine mdrun. Generally that is desirable when running on a multi-node cluster, and necessary when using multi-simulation algorithms. Usually also installing a build of GROMACS configured without MPI is convenient for users.
Should my version be compiled using double precision?

In general, GROMACS only needs to be build in its default mixed-precision mode. For more details, see the discussion in Chapter 2 of the reference manual. Sometimes, usage may also depend on your target system, and should be decided upon according to the individual instructions.

Questions concerning system preparation and preprocessing¶

Where can I find a solvent coordinate file for use with solvate?

Suitable equilibrated boxes of solvent structure files can be found in the $GMXDIR/share/gromacs/top directory. That location will be searched by default by solvate, for example by using -cs spc216.gro as an argument. Other solvent boxes can be prepared by the user as described on the manual page for solvate and elsewhere. Note that suitable topology files will be needed for the solvent boxes to be useful in grompp. These are available for some force fields, and may be found in the respective subfolder of $GMXDIR/share/gromacs/top.
How to prevent solvate from placing waters in undesired places?

Water placement is generally well behaved when solvating proteins, but can be difficult when setting up membrane or micelle simulations. In those cases, waters may be placed in between the alkyl chains of the lipids, leading to problems later during the simulation. You can either remove those waters by hand (and do the accounting for molecule types in the topology file), or set up a local copy of the vdwradii.dat file from the $GMXLIB directory, specific for your project and located in your working directory. In it, you can increase the vdW radius of the atoms, to suppress such interstitial insertions. Recommended e.g. at a common tutorial is the use of 0.375 instead of 0.15.

How do I provide multiple definitions of bonds / dihedrals in a topology?

You can add additional bonded terms beyond those that are normally defined for a residue (e.g. when defining a special ligand) by including additional copies of the respective lines under the [ bonds ], [ pairs ], [ angles ] and [ dihedrals ] sections in the [ moleculetype ] section for your molecule, found either in the itp file or the topology file. This will add those extra terms to the potential energy evaluation, but will not remove the previous ones. So be careful with duplicate entries. Also keep in mind that this does not apply to duplicated entries for [ bondtypes ], [ angletypes ], or [ dihedraltypes ], in force-field definition files, where duplicates overwrite the previous values.
Do I really need a gro file?

The gro file is used in GROMACS as a unified structure file format that can be read by all utilities. The large majority of GROMACS routines can also use other file types such as pdb, with the limitations that no velocities are available in this case. If you need a text-based format with more digits of precision, the g96 format is suitable and supported.
Do I always need to run pdb2gmx when I already produced an itp file elsewhere?

You don’t need to prepare additional files if you already have all itp and top files prepared through other tools.

Examples for those can be found in the System Preparation section of this user guide.
How can I build in missing atoms?

GROMACS has no support for building coordinates of missing non-hydrogen atoms. If your system is missing some part, you will have to add the missing pieces using external programs to avoid the missing atom error. This can be done using programs such as Chimera in combination with Modeller, Swiss PDB Viewer, Maestro. Do not run a simulation that had missing atoms unless you know exactly why it will be stable.
Why is the total charge of my system not an integer like it should be?

In floating point math, real numbers can not be displayed to arbitrary precision (for more on this, see e.g. Wikipedia). This means that very small differences to the final integer value will persist, and GROMACS will not lie to you and round those values up or down. If your charge differs from the integer value by a larger amount, e.g. at least 0.01, this usually means that something went wrong during your system preparation

Questions regarding simulation methodology¶

Should I couple a handful of ions to their own temperature-coupling bath?

No. You need to consider the minimal size of your temperature coupling groups, as explained in Thermostats and more specifically in What not to do, as well as the implementation of your chosen thermostat as described in the reference manual.
Why do my grompp restarts always start from time zero?

You can choose different values for tinit and init-step.
Why can’t I do conjugate gradient minimization with constraints?

Minimization with the conjugate gradient scheme can not be performed with constraints as described in the reference manual, and some additional information on Wikipedia.
How do I hold atoms in place in my energy minimization or simulation?

Groups may be frozen in place using freeze groups (see the reference manual). It is more common to use a set of position restraints, to place penalties on movement of the atoms. Files that control this kind of behaviour can be created using genrestr.
How do I extend a completed a simulation to longer times?

Please see the section on Managing long simulations. You can either prepare a new mdp file, or extend the simulation time in the original tpr file using convert-tpr.
How should I compute a single-point energy?

This is best achieved with the -rerun option to mdrun. See the Re-running a simulation section.

Parameterization and Force Fields¶

I want to simulate a molecule (protein, DNA, etc.) which complexes with various transition metal ions, iron-sulfur clusters, or other exotic species. Parameters for these exotic species aren’t available in force field X. What should I do?

First, you should consider how well MD will actually describe your system (e.g. see some of the recent literature). Many species are infeasible to model without either atomic polarizability, or QM treatments. Then you need to prepare your own set of parameters and add a new residue to your force field of choice. Then you will have to validate that your system behaves in a physical way, before continuing your simulation studies. You could also try to build a more simplified model that does not rely on the complicated additions, as long as it still represents the correct real object in the laboratory.
Should I take parameters from one force field and apply them inside another that is missing them?

NO. Molecules parametrized for a given force field will not behave in a physical manner when interacting with other molecules that have been parametrized according to different standards. If your required molecule is not included in the force field you need to use, you will have to parametrize it yourself according to the methodology of this force field.

Analysis and Visualization¶

Why am I seeing bonds being created when I watch the trajectory?

Most visualization softwares determine the bond status of atoms depending on a set of predefined distances. So the bonding pattern created by them might not be the one defined in your topology file. What matters is the information encoded in there. If the software has read a tpr file, then the information is in reliable agreement with the topology you supplied to grompp.
When visualizing a trajectory from a simulation using PBC, why are there holes or my peptide leaving the simulation box?

Those holes and molecules moving around are just a result of molecules ranging over the box boundaries and wrapping around, and are not a reason for concern. You can fix the visualization using trjconv to prepare the structure for analysis.
Why is my total simulation time not an integer like it should be?

As the simulation time is calculated using floating point arithmetic, rounding errors can occur but are not of concern.