MacroModel Force Field Implementation

All MacroModel force field files contain the authentic parameter set published by the original authors of the force field. In addition to these parameters are other parameters from other sources (e.g., the literature or work at Columbia). Parameters in the force field files are labeled as to their origin (O = original from the force field authors, M = modified from the original values, and A = added from some other source where no original parameter exists). They are also labeled by quality (1 = high quality final value, 2 = tentative value based on more than one experimental or quantum calculation, 3 = crude low quality parameter). Sources of A and M parameters are given at the ends of the lines in the force field files.

The MacroModel implementations of standard force fields differ from the authentic force fields in the following ways:

MM2*

All force field equations are identical with those of authentic MM2 from Allinger [1], with the exception of the following:

  • The electrostatic equation (MM2* uses partial charges and Coulomb’s law, whereas MM2 uses bond dipoles and the Jeans equation).

  • The out-of-plane bending equation (MM2* uses an improper torsion while MM2 uses a pyramidalization distance—the difference being insignificant except for substantially distorted sp2 systems).

  • Handling of conjugation (MM2* uses specific V1-V3 torsional terms for various conjugated systems, whereas MM2 uses an SCF π calculation in uncommon systems).

    • The MM2* force field is a modification of MM2. MM2* includes partial charge electrostatics, explicit parametrization of conjugated systems, and extensive additional parameters. Although MM2* is the best general force field for organic molecules, it has reduced accuracy with molecules possessing multiple polar functional groups. In addition, solvation energies are accurate only with nonpolar solvents (e.g. chloroform).

    MM2* is an all-atom model with lone pairs only on ether and hydroxyl oxygens. MM2* assigns constant dielectric electrostatics (e = 1) by default.

    N.L. Allinger's MM2 Force Field, (J .Am. Chem. Soc. 1977, 99, 8127) is the basis for MM2*. The current parameter set is Allinger's 1987 MM2(87) version. The MacroModel (the computational component of MacroModel) implementation of the MM2 field differs from authentic MM2 as follows:

    1. Electrostatics - whereas the authentic MM2 program uses dipole/dipole electrostatic interactions, we use partial atomic charges that are derived from the MM2 dipoles. The field is also set for a distance-dependent dielectric to mimic polarization effects, but this can be changed to a constant dielectric treatment.

    2. Conjugated Systems - Whereas MM2 uses a pi-electron calculation to determine bond orders, we use specific torsional force constants to reproduce known rotational energies of conjugated systems (1,3-dienes, enones, etc).

    3. Out-of-plane Bending - To maintain the planarity of sp2 atoms, MM2 uses an out-of-plane bending equation that MacroModel approximates by an improper torsion. The differences are insignificant except for highly distorted sp2 systems.

    4. Finally, MacroModel uses a special set of parameters for Lennard-Jones (10,12) hydrogen bonding as used in AMBER.

    The MacroModel treatment of MM2 is fully equivalent to the authentic field with the exception of the above, and additional or justifiably modified parameters. Such new parameters are listed in .mmo output files as 'A' or 'M' parameters. FOR AUTHENTIC MM2 CALCULATIONS, USE THE MM2 PROGRAM THAT IS AVAILABLE FROM THE QUANTUM CHEMISTRY PROGRAM EXCHANGE.

    Any missing parameters will be listed in the MacroModel job .log file calculations. Parameters for MacroModel MM2 calculations are taken from the file 'mm2.fld'.

MM3*

All force field equations are identical to those of authentic MM3 from Allinger [3], except for those differences listed above for MM2.

The MM3* force field is a modification of MM3. The modifications include: partial charge electrostatics, explicit parametrization of conjugated systems, and minimal additional parameters. Hydrogen bonding has not yet been implemented.

The MM3* is an excellent force field for simple, monofunctional organic molecules, but it has poor accuracy with molecules having multiple polar functional groups. Solvation energies are accurate only with nonpolar solvents (e.g. chloroform).

MM3* is an all-atom model that uses no lone electron pairs. Constant dielectric electrostatics (e = 1) are assigned by default.

The MM3* is based on N.L. Allinger's MM3 Force Field (J .Am. Chem. Soc. 1989, 111, 8552). The current parameter set is from Allinger's 1991 version. The MacroModel (the computational component of MacroModel) implementation of MM3 differs from authentic MM3 as follows:

1. Electrostatics - whereas the authentic MM3 program uses dipole/dipole electrostatic interactions, we use partial atomic charges that are derived from the MM3 dipoles. The field is also set for a distance-dependent dielectric to mimic polarization effects but this can be changed to a constant dielectric treatment.

2. Conjugated Systems - Whereas MM3 uses a pi-electron calculation of bond orders, MacroModel uses specific torsional force constants to reproduce known rotational energies of conjugated systems (1,3-dienes, enones, etc).

3. Out-of-plane Bending - To maintain the planarity of sp2 atoms, MM3 uses an out-of-plane bending equation that MacroModel approximates by an improper torsion. The differences are insignificant except for significantly distorted sp2 systems.

4. Finally, MacroModel uses a special set of nonbonded parameters for hydrogen bonding atoms which are compatible with partial charge electrostatics.

The MacroModel treatment of MM3 is fully equivalent to the authentic field with the exception of the above and additional or justifiably modified parameters. Such new parameters are listed in .mmo output files as 'A' or 'M' parameters. FOR AUTHENTIC MM3 CALCULATIONS, USE THE MM3 PROGRAM THAT IS AVAILABLE FROM PROFESSOR ALLINGER.

Any missing parameters will be listed in the MacroModel job .log file calculations. Parameters for MacroModel MM3 calculations are taken from the file 'mm3.fld'.

AMBER*

All force field equations are identical to those of authentic AMBER from Kollman with additional parameters for organic functionality [4, 5]. The MacroModel default for hydrogen bonding uses Kollman’s recent 6,12-Lennard Jones treatment [6] and an improved peptide backbone parameter set [7].

The AMBER* force field is a modification of AMBER (84 and 86). The modifications include: new torsional parameters for peptides, Kollman 6,12-hydrogen bonding function, and additional parameters for organics. This is an all-atom and united-atom field that omits carbon-bound hydrogens but retains hydrogens on heteroatoms. Lone electron pairs are placed on sulfur atoms.

AMBER* is most useful for peptide and nucleic acid modeling, but it can also be used for simple organic molecules. Organics having highly interacting functional groups or heterocycles may need additional parameters. Note that hydrogens will need to be added to the heteroatoms of most structures from the Brookhaven data bank (see Hadd/Hdel in the "Edit" menu).

When using the AMBER* force field, it is necessary to use hydrogens on SP2 carbons. For sterically-hindered systems, choose the all-atom model. For polar molecules, extended nonbonded cutoff distances produce the highest accuracy. AMBER's solvation energies are reliable in both water and chloroform. AMBER* applies constant dielectric electrostatics (e = 1) by default.

AMBER* is based on Kollman's Biopolymer Field (Weiner, S. J.; Kollman, P. A.; Case, D.; Singh, U.C.; Alagona, G.; Profeta, S.; Weiner, P. J. Am. Chem. Soc.1984, 106, 765), but it also contains parameters for the all-atom model described by S. J. Weiner, P. A. Kollman, N. T. Nguyen, and D. A. Case (J. Comput. Chem.1987 , 7, 230). The water in the force field file is from the AMBER publications.

To the best of our knowledge, the MacroModel treatment of AMBER is fully equivalent to the authentic field with the exception of additional or justifiably modified parameters. Such new parameters are listed in .mmo output files as 'A' or 'M' parameters. FOR AN AUTHENTIC AMBER CALCULATION, USE THE AMBER PROGRAM WHICH IS AVAILABLE FROM PROFESSOR KOLLMAN.

Any missing parameters will be listed in the MacroModel job .log file. Parameters for MacroModel AMBER calculations are taken from the file amber.fld.

OPLS*

All force field equations are identical to those of OPLS/AMBER from Jorgensen [9].

The OPLS* force field is a modification of the published OPLS (91) force field. New torsional potentials and minimal additional parameters have been added. This force field is useful for peptides and simple organic molecules. Organics having highly interacting functional groups or heterocycles are likely to need additional parameters. Solvation energies in both water and chloroform are reliable.

OPLS* assigns constant dielectric electrostatics (e = 1) by default. All modeling should be done with constant dielectric electrostatics. This force field cannot be used with distance- dependent electrostatics. Both the all-atom and united-atom models of this field can be used with simple systems. However, sterically hindered systems require the use of the all-atom model, and and polar molecules demand extended nonbonded cutoff distances. In either case, hydrogens should be used on SP2 carbons. No lone pairs are used in the OPLS* force field.

The OPLS* force field is based on W.L. Jorgensen's OPLS force field for peptides. OPLS* includes OPLS charges and nonbonded potentials, in addition to AMBER bonded energy components, as described by W.L. Jorgensen and J. Tirado-Rives (J. Am. Chem. Soc. 1988, 110, 1657). The water in the file is TIP3P (Jorgensen, W. L.; et al. J. Chem. Phys. 1983, 79, 926).

A dielectric constant (set in the Potential tab of the MacroModel panels) of 1.0 is recommended, along with use of the extended distance nonbonded cutoff option. The OPLS* field is an interesting alternative to AMBER, since it uses no explicit hydrogen binding functions or van der Waals parameters for hydrogens that are bound to heteroatoms. Note, however, that OPLS* may produce conformational energy differences that are substantially in error, since many of the bonded parameters have been taken directly from AMBER, without systematic adjustment to offset the OPLS nonbonded parameters. While many of the parameters have been adjusted to better reproduce conformational energy differences but it is suggested that any such conformational calculations be checked against AMBER or MM2.

To the best of our knowledge, the MacroModel treatment of OPLS* is fully equivalent to the authentic force field with the exception of additional or justifiably modified parameters. Such new parameters are listed in .mmo output files as 'A' or 'M' parameters.

Any missing parameters are listed in the MacroModel job .log file calculations. Parameters for MacroModel OPLS calculations are taken from the file oplsa.fld.

OPLS_2005

OPLS_2005 is an enhanced version of the OPLS_2001 all atom force field developed by Schrödinger to provide a larger coverage of organic functionality. While retaining the features of OPLS-AA and OPLS_2001, torsional parameters have been refit to reproduce the conformational energetics derived at a higher level of quantum theory; additional stretch, bend, and torsion parameters and charges have been fit to support additional organic functionality. A much larger data set was analyzed for validation of the force field. The parameters for proteins have been updated to the ones published more recently [53].

Note: The OPLS_2001 force field is now considered obsolete and is no longer available. The following information is retained for its connection with OPLS_2005:

OPLS_2001 (or OPLSAA), developed by Professor W. Jorgensen of Yale University, is probably the best available for condensed-phase simulations of peptides. All force-field equations are identical to those of authentic OPLSAA [17]. Schrödinger’s implementation has been validated by comparison to BOSS OPLSAA calculations for a wide variety of organic systems. Comparisons to ab initio calculations and experiment show that OPLS_2001 reproduces conformational energies well for systems for which it has been specifically parameterized. Especially good results can be expected for proteins. With the exception of improved charge, van der Waals and torsion parameters for sulfur in thiols and thiol ethers [18], all parameters are native OPLS_2001. The new thio parameters, which use appreciably smaller charges on sulfur and which have been validated in liquid-phase simulations on thiols and thiol ethers, significantly improve the conformational energetics of CYS and MET residues in proteins.

OPLS4

The OPLS4 force field extends the OPLS3e [63], OPLS3 [62] and OPLS_2005 force fields to cover a much wider range of chemical space with better accuracy. This force field can be customized by using the Force Field Builder - OPLS4/OPLS5 Panel.

In input and output files, the string OPLS4 is used to refer to the OPLS4 force field. Note that the OPLS4 force field is called S-OPLS in command-line applications.

When using this force field, please use the following citation:

  • OPLS4, Schrödinger, Inc., New York, NY, 2021

In addition, please also cite the following paper:

  • Lu, C.; Wu, C.; Ghoreishi, D.; Chen, W.; Wang, L.; Damm, W.; Ross, G. A.; Dahlgren, M. K.; Russell, E.; Von Bargen, C. D.; Abel, R.; Friesner. R. A.; Harder, E.; OPLS4: Improving Force Field Accuracy on Challenging Regimes of Chemical Space. J. Chem. Theory Comput.2021, 17, 4291. DOI: 10.1021/acs.jctc.1c00302

 

AMBER94

All force field equations and parameters are the same as in Cornell et al. [8], with the following small exceptions:

  • In MacroModel partial charges are specified by bond dipoles rather than as charge values. The partial charges may differ slightly between the two implementations; these differences are typically in the fifth significant figure.

  • The atoms defining improper torsions are not specified by the AMBER protocol in situations of high local symmetry. This may sometimes give rise to small differences in molecular energies or geometries between the two programs.

  • The paper gives the two nitrogen types different van der Waals parameters, but the AMBER 4.1 program uses the same parameters for both. We follow the program’s convention.

AMBER94 is the authentic AMBER 94 force field. It all-atom model that uses no lone pairs. This field works well for biopolymers (peptides, proteins and nucleic acids) but has few parameters for organic molecules. The option for extended nonbonded cutoff distances is recommended to obtain the highest accuracy with polar molecules. AMBER assigns constant dielectric electrostatics (e = 1) by default. Solvation energies of the AMBER field are reliable in both water and chloroform.

This version of AMBER is based on the 1994 parameter set of Kollman et al. (Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.; Merz, K. M.; Ferguson, D. M.; Spellmeyer, D. C.; Fox, T.; Caldwell, J. W.; Kollman, P. A. J. Am. Chem. Soc.1995, 117, 5179.) What follows are some notes regarding the MacroModel (the computational component of MacroModel) implementation of this force field.

  • Unlike AMBER*, the new AMBER parameter set contains only parameters for carbons with all attached hydrogens explicitly represented.

  • AMBER94 is designed for use in condensed phases. For this reason, it is recommended that the GB/SA solvation model be used in conjunction with the AMBER94 parameter set. The charge derivation method used in AMBER94 (RESP fitting to HF/6-31G* wave functions) gives charges that perform well in calculations of absolute solvation free energies of small molecules (see the section "Absolute Free Energies of Hydration from the GB/SA model - Dependence on Partial Charges" in the MacroModel Technical Manual). We therefore anticipate that the combination of the GB/SA model and the AMBER94 parameters will be a powerful methodology for simulations of biomolecules in solution.

  • Unlike AMBER*, AMBER94's parameters have not been extended beyond those reported by Kollman et al. As a result AMBER94's utility with organic functionalities is limited, and it is recommended that it is used only with proteins and nucleic acids.

  • The MacroModel AMBER94 implementation uses the same van der Waals parameters for all both amide and amine nitrogens. Although the Kollman paper gives the two nitrogen types different VDW parameters, we have continued to use the same VDW parameter for both in order to maintain consistency with the AMBER4.1 program.

  • Lone pairs are not used on sulfur (or any other) atoms in AMBER94. Kollman et al. point out that neutral sulfur atoms are rarely seen as hydrogen bond acceptors.

  • AMBER94 as implemented in MacroModel should give results very close to those obtained with the AMBER 4.1 program suite. There are, however, some areas where the results obtained with the two programs may differ slightly. In MacroModel the partial atomic charges are derived from bond dipoles, not stored as absolute values as in AMBER4.1. The charges differ in the fifth decimal place and this electrostatic energies obtained with MacroModel will be slightly different to those obtained with AMBER4.1. This difference is of the order of 1 kJ/mol for a large, highly charged structure such as d(ATGC).d(TACG). Additionally, there may be some small differences between the improper torsion energies calculated with MacroModel and those calculated with AMBER4.1. These arise because, for some generalized improper torsions, the two programs can select different ways to measure the angle. All other energy terms are calculated identically between the two programs.

MMFF

Our implementation is identical to that described by Halgren [10–16]. We supply both MMFF94 and MMFF94s; the latter enforces planarity about delocalized sp2 nitrogens.

Whenever a current energy calculation (ECalc) is carried out, a listing file (jobname.mmo) can be produced which contains all parameters used in the calculation along with the origin and quality of each parameter. Note that any torsion parameter where V1- V3 are all set to zero will not be included in the output. To include these in the listing it is necessary to include the DEBG 56 command in the command file: see Miscellaneous and Debugging Opcodes for details. This automatic parameter referencing feature provides important information on the quality and reliability of the calculation.

Different force fields use different defaults for their electrostatic treatment (constant or distance-dependent dielectric) and their cutoff distances (van der Waals and electrostatic). It is possible to set such options exactly as in the authentic fields using the Electrostatic Treatment and Cutoff option menus in the Potential tab of the MacroModel panels.

The MMFF and MMFFs force fields are all-atom models. These fields are especially useful for biopolymers (peptides and proteins). MMFF's solvation energies are reliable in both water and chloroform. MMFF assigns constant dielectric electrostatics (e = 1) by default. Extended nonbonded cutoff distances should be used for high accuracy with polar molecules. When a conformational search is being used as a precursor to a JBW (IMPS) simulation, the same force-field (MMFF or MMFFs) should be used in both.

The MMFF94 (or MMFF for short) force field was introduced by Thomas A. Halgren (J. Comput. Chem.1996, vol. 17, no. 5). MMFF is an all-atom force field, with no lone pairs. Thus, its MacroModel "H-addition profile" is the same as those of AMBER94 and MM3. MMFF has been carefully parametrized for a wide variety of organic functional groups and for proteins (see MMFF-I). It should thus be well suited for drug-receptor studies. MMFF charges (see MMFF-II) are derived in a manner suitable for use with the GB/SA solvation model.

There are actually two MMFF force-fields implemented: MMFF and MMFFs, where the "s" is for "static." MMFF gives slightly pyramidal amide geometries and rather strongly pyramidal aniline, nucleic-acid base, and other enamine nitrogens. Experimental and theoretical studies indicate that this reflects the true ground- state geometry.

In contrast, MMFFs uses modified torsion and out-of-plane bending parameters for these nitrogens, giving rise to planar or nearly planar geometries. Such nitrogens appear planar in crystal structures, due to time-averaging of the rapidly inverting pyramidal geometries.

Regular MMFF likely is most useful for dynamics studies (where actual time averaging yields planar average structures), whereas the MMFFs is probably most useful for conformational search and for minimizations that are to be compared with crystallographic data. With the obvious exception of inversion barriers, the two force fields give similar (and, when delocalized trigonal nitrogens are not involved, identical) relative conformational energies.

MMFF treats "protein metals" such as Zn+2, Cu+, Cu+2, Fe+2, and Fe+3, as well as an assortment of alkali and alkaline earth cations, via the commonly used "nonbonded" model, which employs van der Waals and electrostatic interactions between the ligand and metal atoms. Covalent systems involving metals can be accommodated by means of user-defined atom types and parameters.

The MacroModel implementation of MMFF was developed in cooperation with Dr. Halgren. And unlike our other force fields, the assignment of parameters to interactions (stretch, bend, nonbonded, etc.) is handled by a server co-process, called "mmff_setup," which was written by Dr. Halgren. But, as with any other force-field, MMFF parameters can be printed to the .mmo file using the Maestro Current Energy panel with complete listing (MacroModel "ELST 1") and displayed using the force field viewer (see Force Field Viewer on the Tools menu).

Communication between the server process and Maestro is handled by an interface layer called the BMFF library. For details of how BMFF works and for a discussion of how to add new or overriding parameters to the MMFF force-field file (f10.fld), see The BMFF Protocol.