AMBER force field ports for the GROMACS molecular dynamics suite

Maintained by Eric J. Sorin, Ph.D.
Department of Chemistry & Biochemistry
California State University, Long Beach

With significant contributions from:
 Sanghyun Park Ph.D., Jory Z. Ruscio Ph.D., Erik Thompson*, and Kristin Haldeman*

(*CSULB undergraduate students)

Are you using our ffamber ports for an upcoming publication? If so, please cite:
Sorin & Pande (2005), Biophysical Journal, 88, 2472-2493.
DePaul, Thompson, Patel, Haldeman, & Sorin (2010), Nucleic Acids Research, 38, 4856-4867.

AMBER homepage
General Information
Force Field Information
Implementation & Validation
Installation & Testing
! Important notes on using the AMBER ports!
Contact Us
GROMACS homepage

Update: Our ffAMBER ports are now part of newer GROMACS distributions. ACPYPE - AnteChamber PYthon Parser interfacE should be useful to those who wish to add molecular parameters, import AMBER topologies, etc.

General Information:

We have ported the following AMBER potentials and TIP water models for use in the GROMACS MD suite. AMBER ports for GROMACS versions 3.1.4, 3.2.1, and 3.3/3.3.1 have been tested against AMBER 8.0, as discussed below. As there have been several versions of the TIP parameters published, we have taken them from the most recent paper of the Jorgensen group to describe TIP water (Mahoney, 2000). In addition to these force fields, we have added several ions, the urea molecule (URE), hydroxyproline (HYP), ornithine (ORN), and diaminobutyric acid (DAB), as described below. Please cite the following tabulated citations when using any of these force fields.

AMBER conventions have been maintained where possible. However, due to specific GROMACS functionality this was not always feasible. As detailed in this page, there are several very important caveats to using these AMBER ports properly within the GROMACS suite. Please read the following information carefully before using these AMBER ports, and refer also to the ffAMBER FAQ if you have questions that are not answered below. This index.html and supporting files can be found in the README subdirectory of the ffamber*tar.gz downloads, which are available with or without the pdf documentation listed in the table below.

Due to the caveats mentioned above, ffamber ports are not yet distributed as part of the GROMACS package. We are currently working with GROMACS developers Erik Lindahl & David van der Spoel to make the necessary changes in GROMACS that will allow us to fully merge these ports with upcoming GROMACS distributions and thereby simplify the use of AMBER in GROMACS.

Potential Port Name Literature DL
AMBER-94 ffamber94 Cornell et al. (1995), JACS 117, 5179-5197 PDF
AMBER-96 ffamber96 Kollman (1996), Acc. Chem. Res. 29, 461-469 PDF
AMBER-99 ffamber99 Wang et al. (2000), J. Comp. Chem. 21, 1049-1074 PDF
AMBER-03* ffamber03 Duan et al. (2003), J. Comp. Chem. 24, 1999-2012 PDF
AMBER-GS ffamberGS Garcia & Sanbonmatsu (2002), PNAS 99, 2782-2787 PDF
AMBER-GS-S** ffamberGSs Nymeyer & Garcia (2003) PNAS 100, 13934-13939 PDF
AMBER-99f*** ffamber99p Sorin & Pande (2005). Biophys. J. 88(4), 2472-2493 PDF
AMBER-99SB ffamber99sb Hornak et. al (2006). Proteins 65, 712-725 PDF
TIP3P HOH or T3P Mahoney & Jorgensen (2000). J. Chem. Phys. 112, 8910-8922 PDF
TIP3P(heavy)**** T3H Mahoney & Jorgensen (2000). J. Chem. Phys. 112, 8910-8922 PDF
TIP4P T4P Mahoney & Jorgensen (2000). J. Chem. Phys. 112, 8910-8922 PDF
TIP4P-Ew T4E Horn et al. (2004). J. Chem. Phys.120, 9665-9678 PDF
TIP5P T5P Mahoney & Jorgensen (2000). J. Chem. Phys. 112, 8910-8922 PDF
* all-atom only, does NOT include the AMBER-03ua united-atom potential
** no 1-4 vdW scaling
*** helix-coil force field
**** GROMACS heavy H2O definition using TIP3P parameters

Force Field Information:

AMBER-94 is one of the most widely used and understood of the seminal all-atom force fields, though known to be somewhat "helix-friendly."

AMBER-96 is a variant of AMBER-94 in which peptide f/y torsional potentials have been adjusted to yeild better agreement between molecular mechanical and quantum mechanical energetics (disfavoring helices, favoring extended conformations).

AMBER-GS is identical to AMBER-94 with peptide f/y torsional potentials removed (set to zero). This AMBER-GS differs from Angel Garcia's force field with a minor modification: Garcia's version also does *NOT* scale 1-4 vdW interactions (personal comm., A. Garcia), as is standard for all AMBER force fields. Please note this distinction when choosing between these force fields [for more information, see Sorin & Pande (2005). J. Comp. Chem. 26 (7), 682-690]. To use the "true AMBER-GS" force field that removes 1-4 vdW scaling, which we have referred to as AMBER-GS-S, please use the ffamberGSs files provided in our ffAMBER ports.

AMBER-99 is the "3rd generation" update to AMBER-94, including updated parameters for both amino and nucleic acids. The primary changes for peptides/proteins is again in the f/y torsional potentials (again to disfavor helical conformations).

AMBER-99f is our AMBER-99 variant for helix-coil simulations and replaces the f torsion in AMBER-99 with that of the Cornell (AMBER-94) potential.

AMBER-99SB is the AMBER-99 variant developed by Simmerling and co-workers with updated backbone f and y torsions fit to ab initio calculations (ported by Jory Z. Ruscio).

AMBER-03 is a variant of the AMBER-99 potential in which charges and main-chain torsion potentials have been rederived based on QM+continuum solvent calculations and each amino acid is allowed unique main-chain charges (nucleic acids have not been modified from AMBER-99).

For more information on each of these FF's, please refer to the AMBER homepage and the references above.

The following additions have also been made:

Ornithine (ORN) and diaminobutyric acid (DAB) residues were defined using standard AMBER-94 and AMBER-99 Lennard-Jones and bonded parameters. ORN sidechain charges were taken from the AMBER-99 port in TINKER 4.0; DAB sidechain charges were fit to maintain proper LYS --> ORN --> DAB heavy atom charge group trends, as described in:
B. Zagrovic, J. Lipfert, E.J. Sorin, I.S. Millett, W.F. Van Gunsteren, S. Doniach, & V.S. Pande (2005) PNAS, 102(33), 11698-11703.

The hydroxyproline (HYP) residue has also been added* in accord to the following publications:
S.D. Mooney, P.A. Kollman & T.E. Klein (2002), Biopolymers 64, 63-71.
S. Park, R.J. Radmer, T.E. Klein & V.S. Pande (2005), J. Comp. Chem. 26, 1612–1616.
*due to the significant force field differences, HYP is not present in AMBER-03 ports

The urea molecule (URE) was defined using standard AMBER Lennard-Jones and bonded parameters, and RESP charges were derived by Jim Caldwell.

The AMBER-94 and AMBER-99 force fields were ported and validated first. Force field files for the variants of AMBER-94 and AMBER-99 were then generated after validation by making the minor modifications required (i.e. peptide f/y and 1-4 vdW scaling).

While newer GROMACS distributions support multiple torsional terms for a given atomic quartet, v3.1.4 did not. Proper torsion potentials were therefore converted to Ryckaert-Bellemans series in a fashion similar to that described in the GROMACS manual for OPLS torsions. Additionally, improper torsions are handled in GROMACS using the proper dihedral definition to match the AMBER standard, rather than the default analytical function used for improper dihedrals in GROMACS. Because we found inconsistencies in the definitions of improper torsions when using the AMBER 8.0 package (i.e. ordering of the atomic quartet), impropers in our ports were set to agree with AMBER 8.0, and proper torsions were then checked for exact agreement between all AMBER 8.0 Fourier series torsions and the GROMACS Ryckaert-Bellemans series torsions.

To validate the potential energetics of our ports, structures for all monomer, homodimer, and homotrimer systems (including amino and nucleic acids) were generated using the LEAP program in AMBER and these structures were perturbed by adding a random number between -0.2 and +0.2 Angstroms to all 3N coordinates. This step is necessary because default structures generated by AMBER's LEAP program contain many low energy terms which are not suitable for quantitative force field validation. All potential energy components were compared between GROMACS 3.1.4/3.2.1/3.3/3.3.1 and AMBER 8.0 and the resulting energy differences were generally < 0.005%. These results were not significantly altered by taking the mean over 10 independent perturbations for each monomer/dimer/trimer system or altering the magnitude of the applied 3N perturbations.

Mean relative errors for all
monomers, homodimers, and homotrimers*
(using single precision binaries)
<|Error|> (%) AMBER-94 AMBER-99 AMBER-03
Amino Nucleic Amino Nucleic Amino Nucleic**
BOND 0.000 0.000 0.000 0.000 0.000 0.000
ANGLE 0.000 0.000 0.000 0.000 0.000 0.000
DIHEDRAL 0.004 0.000 0.003 0.000 0.005 0.000
1-4 vdW 0.000 0.001 0.001 0.001 0.000 0.001
1-4 QQ 0.001 0.001 0.001 0.001 0.001 0.001
vdW 0.001 0.002 0.001 0.006 0.001 0.006
QQ 0.003 0.004 0.003 0.003 0.003 0.003
Total Potential 0.001 0.001 0.001 0.001 0.001 0.001
*highest mean error from GROMACS 3.1.4, 3.2.1, 3.3/3.3.1
**unchanged from AMBER-99, tested for consistency

(1) Install the desired GROMACS distribution (v3.1.4, v3.2.1, v3.3, or v3.3.1).
(2) Download the appropriate ffamber ports (.tar.gz) with or without pdf documentation from the table below, being sure that the version number you choose matches the version of GROMACS you are using.
(3) Unzip/untar the downloaded tar.gz file.
(4) Copy aminoacids.dat and vdwradii.dat to the "top" directory in your gromacs distribution (you should see force field files there, such as ffoplsaa.*). If you plan on simulating nucleic acids, refer to the note for nucleic acids in aminoacids.dat below.
(5) Files for each force field are located in a seperate subdirectory, such as ffamber94/ for the Cornell potential. Copy the desired ffamber* files to the top directory in your gromacs distribution.
(6) Increment the number at the top of the "top/FF.dat" file by 1 for each AMBER port you'll install (so that it matches the total number of forcefields available in the "top" directory).
(7) Add lines like the following to the "top/FF.dat" file. These are used by pdb2gmx to allow you to identify the desired FF and field 1 must match the ffamber* filename prefixes, whereas the following fields can be user-defined:
     ffamber94 AMBER94 Cornell protein/nucleic forcefield
     ffamber99 AMBER99 Wang protein/nucleic acid forcefield
     ffamber99p AMBER99p protein/nucleic forcefield
     ffamber03 AMBER03 Duan protein/nucleic forcefield
(8) Locate the GMXRC in your GROMACS distribution and run `source GMXRC`.
(9) Run `pdb2gmx -H14 -f any.pdb` with any pdb to verify that these force fields are now seen by GROMACS. Working example .pdb files are available below, alongside pre-prepared gro and top files (GROMACS 3.1.4 / AMBER94) to which you can compare your resulting files.

Protein DNA RNA
Villin headpeice Dickerson dodecamer GCAA tetraloop hairpin
AMBER94 gro file AMBER94 gro file AMBER94 gro file
AMBER94 top file AMBER94 top file AMBER94 top file

(10) Modify residue names in your pdb file(s) as described below to properly generate .gro and .top files using pdb2gmx (the files above are ready to use).

AMBER port distributions
Version Original
Release Date
Download Notes on this release
3.1.4 July 2005 with pdfs
without pdfs
Must use -H14 flag in pdb2gmx
3.2.1 July 2005 with pdfs
without pdfs
-H14 flag no longer necessary for pdb2gmx
3.3 Nov. 2005 with pdfs
without pdfs
.hdb file format change,
hydrogen database now functional for nucleic acids
3.3.1 May 2006 with pdfs
without pdfs
This port has also been validated using GMX 3.3.3,
but may give warning or error messages using GMX 3.3.2
4.0 Feb 2009 with pdfs
without pdfs
This port has been validated using GMX 4.0.2, 4.0.3, 4.0.4 & 4.0.7
NOTE: pdb2gmx auto-converts residue CYM to CYN in 4.0.2 & 4.0.3

ffAMBER tar.gz files are ~7.5 MB with .pdfs and ~850 kB without

This list includes only the most important information needed to properly use our AMBER ports inside the GROMACS suite. Non-vital information is compiled in the ffAMBER FAQ listed above. Information on using GROMACS as well as FAQs, tutorials, and a user forum are available at http://www.gromacs.org/.

(!) -H14 flag in gromacs: You must use the -H14 flag when running the GROMACS v3.1.4 version of pdb2gmx to get *ALL* hydrogen-hydrogen 1-4 interactions that should be present in AMBER topologies. This has been corrected in v3.2.1 and the -H14 flag is no longer necessary.
(!) Residue Nomenclature: Residues in the AMBER ports are named according to their position in the sequence (i.e. terminal, non-terminal, monomer) following standard AMBER conventions. For this reason, it may be necessary to rename residues in the .pdb file you will import beforehand. Please note that all residues are named in the residue topology files (i.e. ffamber*.rtp), so if you are unsure of the correct residue name to use, you should be able to find it there. The .rtp files are ordered as follows: water models (TIP), ions & urea (URE), peptide terminal capping residues (ie. ACE, NH2, NMe), non-terminal amino acids (i.e. TYR, ALA), non-terminal acidic amino acids (i.e. ASH, GLH, etc.), C-terminal amino acids (i.e. CALA, CGLY), N-terminal amino acids (i.e. NALA, NGLY), and all nucleic acid residues. Nucleic acids listed at the end of each .rtp follow the following order for each residue type: DNA is first, followed by RNA, in the order 5'-term, 3'-term, non-terminal, and monomer. The three .pdb files above are examples of how pdb files shoud be modified. Residues in the ffamber ports have been named as follows:
Non-terminal amino and nucleic acid residues follow standard AMBER naming conventions. To avoid confusion between GROMACS and AMBER conventions, we have omitted the redundant HIS residue, leaving HID, HIE, HIP, and terminal versions of these topologies. Additionally, due to the automated changing of certain residue names by pdb2gmx, the LYS and CYS residues have been renamed LYP (Lysine plus) and CYN (Cysteine neutral, compared to AMBER residue CYM = Cysteine minus).
C- and N-terminal amino acids include a C or N prefix respectively, so C-terminal ALA is CALA and N-ternimal PHE is NPHE. As with non-terminal versions, the LYS and CYS terminal residues are listed as NLYP,CLYP and NCYN,CCYN.
Nucleic acid residues come in four flavors. All residue names include XY, where X = D or R for DNA or RNA respectively, and Y = first letter of the nucleotide name. A suffix (XYZ) is added for monomers (Z=N), 5'-terminal (Z=5), and 3'-terminal (Z=3) residues. For example, 3'-term DNA Cytosine = "DC3", 5'-term RNA Cytosine = "RC5", non-terminal DNA Cytosine = "DC", and lone RNA Cytosine monomer = "RCN".
Cys disulfide bonds can be instituted by changing the names of the CYS residues involved in the disulfide bonds to CYS2.
(!) Atom Nomenclature: Once you have modified the residue names in your pdb file according to the rules above, pdb2gmx may report a fatal error like "Fatal error: Atom AA in residue XYZ not found in rtp entry with NN atoms while sorting atoms." This may or may not be caused by the use of an aminoacids.dat file in the "top" directory of your GROMACS distribution that does not include the name of the residue that is causing the problem (XYZ in this case). This can be easily fixed by modifying the name of the problematic atom in your .pdb file to match what is shown for residue XYZ in the .rtp file.
(!) Water nomenclature: residues in pre-solvated pdb files should be named as shown here to use the correct water model in the desired amber port: TIP3P ("HOH" or "T3P"), heavy TIP3P ("T3H"), TIP4P ("T4P"), TIP4P-Ew ("T4E"), and TIP5P ("T5P"). This usage allows any model to be used without modifying the force field (.rtp) files. Note, however, that for the default "HOH" listing the TIP3P model will be is assumed. Also note that because GROMACS supports SOL=HOH water definitions, importing a solvated PDB with names other than HOH or SOL may cause the solvent molecules to be treated as part of the biopolymer rather than listing SOL molecules at the end of the topology, which has to be modified by hand after the fact.
(!) Specifying water models in .top files: If not using the default SPC water model in GROMACS, make sure you list the correct ffamber_tip*.itp file in your gromacs topology file! TIP3P heavy water (H's four times heavier, difference subtracted from oxygen) is defined in ffamber_tip3p_heavy.itp, and the TIP4P-Ewald model uses ffamber_tip4pEW.itp. All other models are specified in the ffamber_tipXp.itp files, where X = {3,4,5}. Water boxes are also present, with the .gro file extension.
(!) For nucleic acid simulations: Amino acid residues have been added to aminoacids.dat, which (i) allows easy use of GROMACS analysis tools and (ii) aids GROMACS in matching atoms in a pdb with those in the residue topology (.rtp) file. However, for nucleic acids this also often causes pdb2gmx to replace an H atom in the first residue of all nucleic acid chains with an incorrect H atom, resulting in non-neutral charge. The correct atom is generally replaced with an atom of type amberXX_25 (hydroxyl H), as pdb2gmx treats it as a terminal hydrogen. For this reason, we provide both the standard aminoacids.dat (without nucleic acids) and a second version with nucleic acid residues included (aminoacids-NA.dat). You can run gcaa.pdb with both versions of aminoacids.dat by copying either file into your "top" directory before each trial and diff'ing the resulting top files to see this difference. There are three ways to handle this:
(a) Include all nucleic acid names in aminoacids.dat and correct the [ atoms ] section of the resulting .top file by hand/script after running pdb2gmx (both the atom type and the charge).
(b) Remove nucleotide names from aminoacids.dat before importing your pdb file and add them back afterward to use standard GROMACS tools.
(c) Do not include nucleic acid names in aminoacids.dat and make an index file which specifies the DNA/RNA atoms to use standard GROMACS tools.
Note that if you choose options (b) or (c) above, you may need to modify the atom names in your .pdb file to agree with the residue listings in the .rtp file due to issue (3) above.
(!) Using AMBER ions: Our AMBER ports include common ion definitions, which are listed in the ffamber*.rtp files (just below the TIP water models). This allows the AMBER ports to be used without modification or use of the GROMACS ions.itp file. At the moment these include Cl- , IB+, Na+, K+ , Rb+, Cs+, Li+ , Ca2+, Mg2+, Zn2+, Sr2+, and Ba2+. If your pdb file has ions present and pdb2gmx does not properly convert those ions, please check the atom and residue name of your ions and rename them if necessary to agree with the .rtp file. If you are using ion-related GROMACS tools, such as genion, you will need to enter the AMBER ion definition to the ions.itp file in the "top" directory of the GROMACS distribution.

Being familiar with the AMBER family of force fields is not necessary for the use of our ports, but it is highly recommended. More importantly, we expect that those using these ports will be very experienced in using both GROMACS and UNIX/LINUX. Please do not contact us with GROMACS/UNIX/LINUX questions. We have done our best to provide very complete documentation above to help GROMACS users with the installation and use of our AMBER ports. Please do not email with implementation questions unless you are confident that your problem/question has not yet been addressed above or in the associated ffAMBER FAQ.

Communications not answered above or on our ffAMBER FAQ should be directed to Eric J. Sorin