Force Fields Guide¶

Force fields are the mathematical models that describe atomic interactions in molecular dynamics. Choosing the right force field combination is critical for simulation accuracy. PRISM supports multiple force field families for both proteins and ligands.

Quick Start

# Default (GAFF + amber99sb)
prism protein.pdb ligand.mol2 -o output

# High accuracy (OpenFF + AMBER14SB)
prism protein.pdb ligand.sdf -o output --ligand-forcefield openff --forcefield amber14sb

Quick Selection Guide¶

Recommended Combinations¶

Use Case	Protein FF	Ligand FF	Water	Command
General Drug Discovery	AMBER14SB	GAFF2	TIP3P	`--forcefield amber14sb --ligand-forcefield gaff2`
High Accuracy	AMBER14SB	OpenFF	TIP3P	`--forcefield amber14sb --ligand-forcefield openff`
CHARMM Ecosystem	CHARMM36	CGenFF	TIPS3P	`--forcefield charmm36 --ligand-forcefield cgenff`
Literature Match	AMBER99SB-ILDN	GAFF	TIP3P	`--forcefield amber99sb-ildn --ligand-forcefield gaff`
Small Molecules	OPLS-AA	OPLS-AA	TIP4P	`--forcefield oplsaa --ligand-forcefield opls`

Quick Examples¶

# Default (fast setup, good results)
prism protein.pdb ligand.mol2 -o output

# High accuracy (modern force fields)
prism protein.pdb ligand.sdf -o output \
  --forcefield amber14sb \
  --ligand-forcefield openff \
  --water tip3p

# CHARMM-based workflow
prism protein.pdb ligand.mol2 -o output \
  --forcefield charmm36 \
  --ligand-forcefield cgenff \
  --forcefield-path /path/to/cgenff_files \
  --water tips3p

Protein Force Fields¶

AMBER Family (Recommended)¶

The AMBER force fields are the most widely used for biomolecular simulations.

AMBER99SB¶

prism protein.pdb ligand.mol2 -o output --forcefield amber99sb

- Year: 2006 - Best for: General protein simulations, broad compatibility - Strengths: Well-tested, stable, large validation dataset - Limitations: Some α-helix bias, limited side chain sampling - Water: TIP3P (default)

AMBER99SB-ILDN¶

prism protein.pdb ligand.mol2 -o output --forcefield amber99sb-ildn

- Year: 2010 - Best for: Improved side chain dynamics - Strengths: Better χ1/χ2 rotamer distributions, reduced helix bias - Limitations: Still uses AMBER99SB backbone - Water: TIP3P - Citation: Lindorff-Larsen et al., Proteins (2010)

AMBER14SB ⭐ Recommended¶

prism protein.pdb ligand.mol2 -o output --forcefield amber14sb

- Year: 2015 - Best for: Most modern applications, drug discovery - Strengths: - Improved backbone and side chain parameters - Better agreement with NMR data - Fixed α-helix overstabilization - Water: TIP3P - Citation: Maier et al., J. Chem. Theory Comput. (2015)

AMBER03¶

prism protein.pdb ligand.mol2 -o output --forcefield amber03

- Year: 2003 - Best for: Specific compatibility with older studies - Note: Generally superseded by newer versions

CHARMM Family¶

CHARMM force fields excel for membrane proteins and lipid systems.

CHARMM36¶

prism protein.pdb ligand.mol2 -o output \
  --forcefield charmm36 \
  --water tips3p

- Year: 2012-2017 - Best for: Membrane proteins, protein-lipid interactions - Strengths: Excellent lipid parameters, good protein conformations - Water: TIPS3P (modified TIP3P for CHARMM) - Ligand: Use CGenFF for consistency - Citation: Best et al., J. Chem. Theory Comput. (2012)

CHARMM27¶

prism protein.pdb ligand.mol2 -o output --forcefield charmm27

- Year: 2004 - Note: Superseded by CHARMM36

OPLS Family¶

OPLS-AA excels for small molecule interactions.

OPLS-AA/L¶

prism protein.pdb ligand.mol2 -o output \
  --forcefield oplsaa \
  --water tip4p

- Year: 2001-2005 - Best for: Small molecule interactions, organic chemistry - Strengths: Excellent liquid properties, validated against QM - Water: TIP4P (better with OPLS) - Ligand: Use OPLS-AA ligand FF for consistency - Citation: Jorgensen et al., J. Am. Chem. Soc. (1996)

GROMOS Family¶

prism protein.pdb ligand.mol2 -o output \
  --forcefield gromos96 \
  --water spc

- Best for: United-atom simulations, faster calculations - Note: United-atom (no explicit hydrogens on carbons) - Water: SPC or SPC/E

Ligand Force Fields (8+ Options)¶

PRISM supports 8+ ligand force field generators. Choose based on accuracy needs, compatibility, and chemical coverage.

Comparison Table¶

Force Field	Method	Charge Model	Coverage	Speed	Accuracy	Requires
GAFF	Rule-based	AM1-BCC	Excellent	Fast	Good	AmberTools
GAFF2	Improved rules	AM1-BCC	Excellent	Fast	Better	AmberTools
OpenFF	Data-driven	AM1-BCC-ELF10	Very Good	Medium	Excellent	OpenFF
CGenFF	CHARMM-based	CGenFF server	Good	Medium	Good	Web download
OPLS-AA	Server	CM1A/CM5	Good	Medium	Very Good	Internet
MMFF	SwissParam	MMFF94	Good	Fast	Fair	Internet
MATCH	SwissParam	MATCH	Good	Fast	Fair	Internet
Hybrid	SwissParam	Mixed	Good	Fast	Fair	Internet

1. GAFF (General AMBER Force Field)¶

Default choice - fast and reliable

# Using GAFF
prism protein.pdb ligand.mol2 -o output --ligand-forcefield gaff

Details: - Method: Rule-based atom typing - Charges: AM1-BCC (semi-empirical) - Coverage: Excellent - most organic molecules - Speed: Fast (~1 minute) - Accuracy: Good for most applications - Input: MOL2 (with correct atom types) - Requirements: AmberTools, ACPYPE

Best for: - Rapid prototyping - High-throughput screening - Standard drug-like molecules

Workflow:

ligand.mol2 → antechamber (AM1-BCC) → parmchk2 → ACPYPE → GROMACS

2. GAFF2 (Improved GAFF)¶

Recommended over GAFF for new projects

# Using GAFF2
prism protein.pdb ligand.mol2 -o output --ligand-forcefield gaff2

Details: - Method: Improved rule-based parameters - Charges: AM1-BCC - Improvements: Better torsion parameters, expanded coverage - Speed: Same as GAFF - Accuracy: ~10-15% improvement over GAFF

Best for: - Modern drug discovery projects - When GAFF is too inaccurate but OpenFF is unavailable - Compatibility with AMBER ecosystems

Advantages over GAFF: - Better torsional barriers - Improved bond/angle parameters - Validated against larger dataset

3. OpenFF (Open Force Field) ⭐¶

Most accurate modern force field for small molecules

# Using OpenFF
prism protein.pdb ligand.sdf -o output --ligand-forcefield openff

Details: - Method: Data-driven, fitted to QM calculations - Charges: AM1-BCC-ELF10 or RESP - Coverage: Very good organic molecules - Speed: Medium (~2-5 minutes) - Accuracy: Excellent - best for binding free energy - Input: SDF (preferred), MOL2 - Requirements: OpenFF Toolkit, OpenFF Interchange

Best for: - High-accuracy studies - Binding free energy calculations - Novel chemical scaffolds - Publication-quality results

Workflow:

ligand.sdf → OpenFF Toolkit → Parameter assignment → Interchange → GROMACS

Advantages: - Systematic optimization against QM - Regular updates and improvements - Explicit uncertainty quantification - Best choice for FEP/TI calculations

4. CGenFF (CHARMM General Force Field)¶

For CHARMM-based workflows

# Step 1: Upload ligand to https://cgenff.umaryland.edu/
# Step 2: Download parameter files to /path/to/cgenff/

# Using CGenFF
prism protein.pdb ligand.mol2 -o output \
  --ligand-forcefield cgenff \
  --forcefield-path /path/to/cgenff_files

Details: - Method: Web-based parameterization - Charges: CGenFF server assignment - Coverage: Good for drug-like molecules - Speed: Manual (web upload) - Accuracy: Good, especially with CHARMM protein FF - Requirements: Web-downloaded files

Best for: - CHARMM36 protein force field users - Membrane protein simulations - Consistency with CHARMM-GUI workflows

Workflow:

Upload to ParamChem → Download .str file → PRISM converts → GROMACS

Note: Requires manual web interaction, not suitable for automation

5. OPLS-AA (via LigParGen)¶

Excellent for liquid properties

# Using OPLS-AA
prism protein.pdb ligand.mol2 -o output --ligand-forcefield opls

Details: - Method: LigParGen web server API - Charges: CM1A (or CM5 for higher accuracy) - Coverage: Good for organic molecules - Speed: Medium (~2-3 minutes, requires internet) - Accuracy: Very good for small molecules - Requirements: Internet connection, RDKit

Best for: - OPLS-AA protein force field users - Small molecule interactions - Liquid-phase properties - Organic chemistry simulations

Workflow:

ligand.mol2 → LigParGen API → Server processing → Download → GROMACS

Advantages: - Well-validated for organic liquids - Consistent with OPLS protein FF - Good solvation properties

6-8. SwissParam Force Fields (MMFF, MATCH, Hybrid)¶

Quick parameterization via SwissParam server

# MMFF-based
prism protein.pdb ligand.mol2 -o output --ligand-forcefield mmff

# MATCH-based
prism protein.pdb ligand.mol2 -o output --ligand-forcefield match

# Hybrid MMFF/MATCH
prism protein.pdb ligand.mol2 -o output --ligand-forcefield hybrid

Details: - Method: SwissParam web server API - Charges: MMFF94 or MATCH - Coverage: Good for most organics - Speed: Fast (~1-2 minutes, requires internet) - Accuracy: Fair to good - Requirements: Internet connection

MMFF (Merck Molecular Force Field): - Classical pharmaceutical force field - Good for drug-like molecules - Fast parameterization

MATCH: - More modern parameterization - Better coverage of functional groups

Hybrid: - Combines MMFF and MATCH - Aims for best of both

Best for: - Quick testing - Preliminary simulations - When other methods fail - Exotic chemistry

Limitations: - Lower accuracy than OpenFF - Less validation than GAFF - Not ideal for quantitative studies

Gaussian RESP Charges¶

For the highest-quality charge assignments, PRISM can use Gaussian to compute RESP charges instead of the default AM1-BCC method:

# HF/6-31G* RESP charges (standard for AMBER)
prism protein.pdb ligand.mol2 -o output --gaussian hf

# DFT-level RESP charges (B3LYP)
prism protein.pdb ligand.mol2 -o output --gaussian dft

# With geometry optimization before charge calculation
prism protein.pdb ligand.mol2 -o output --gaussian hf --isopt opt

# From a pre-computed RESP file
prism protein.pdb ligand.mol2 -o output --respfile charges.resp

Flag	Description	Default
`--gaussian` / `-g`	Gaussian method (`hf` or `dft`)	off
`--isopt`	Optimization level before RESP	none
`--respfile` / `-rf`	Path to pre-computed RESP file	none
`--nproc`	Gaussian CPU cores	`16`
`--mem`	Gaussian memory	`4GB`

Gaussian Required

This feature requires a working Gaussian installation accessible from the command line. The resulting RESP charges replace the default AM1-BCC charges for GAFF/GAFF2.

Theory and Workflow¶

The default AM1-BCC method uses the semi-empirical AM1 Hamiltonian to compute Mulliken charges, then applies bond charge corrections (BCC) to approximate HF/6-31G* electrostatic potentials. While fast, this approach can be inaccurate for molecules with unusual electronic structure. RESP charges from ab initio calculations provide a more rigorous alternative.

Step 1: Geometry Optimization (Optional)¶

When --isopt opt is specified, PRISM first optimizes the ligand geometry at the chosen level of theory:

HF mode: HF/6-31G* (Hartree-Fock, standard for AMBER force fields)
DFT mode: B3LYP/6-31G* (hybrid density functional, better for conjugated systems)

This ensures the charge calculation is performed on a quantum-mechanically consistent geometry rather than on the crystallographic or docked pose.

Step 2: Electrostatic Potential (ESP) Calculation¶

Gaussian computes the molecular electrostatic potential (ESP) on a grid of points surrounding the molecule using the Merz-Kollman scheme:

\[ V_{\text{ESP}}(\mathbf{r}) = \sum_A \frac{Z_A}{|\mathbf{r} - \mathbf{R}_A|} - \int \frac{\rho(\mathbf{r}')}{|\mathbf{r} - \mathbf{r}'|} d\mathbf{r}' \]

where \(Z_A\) and \(\mathbf{R}_A\) are nuclear charges and positions, and \(\rho(\mathbf{r}')\) is the electron density from the HF or DFT wavefunction. The ESP is evaluated at multiple layers of points on van der Waals surfaces (controlled by IOp(6/33=2,6/42=6) in the Gaussian route section).

Step 3: RESP Charge Fitting¶

The Restrained Electrostatic Potential (RESP) method fits atom-centered point charges \(\{q_i\}\) to reproduce the quantum-mechanical ESP. The objective function minimizes:

\[ \chi^2_{\text{RESP}} = \sum_{k=1}^{K} \left[ V_{\text{ESP}}(\mathbf{r}_k) - \sum_{i=1}^{M} \frac{q_i}{|\mathbf{r}_k - \mathbf{R}_i|} \right]^2 + a \sum_{i=1}^{M} \left( q_i^2 - b^2 \right) \]

subject to the constraint:

\[ \sum_{i=1}^{M} q_i = Q_{\text{total}} \]

where \(K\) is the number of ESP grid points, \(M\) is the number of atoms, \(Q_{\text{total}}\) is the net molecular charge, and \(a\) is the restraint weight (default \(a = 0.0005\) for stage 1, \(a = 0.001\) for stage 2). The hyperbolic restraint penalty \(a(q_i^2 - b^2)\) drives chemically buried atoms toward zero charge, preventing unphysical charge buildup on atoms poorly constrained by the ESP grid.

RESP fitting is performed by antechamber (AmberTools) in a standard two-stage procedure, and the resulting charges are automatically substituted into the GROMACS .itp topology file, replacing the default AM1-BCC values.

Complete Workflow¶

graph LR
    A[Ligand MOL2] --> B[Geometry Optimization]
    B --> C[ESP Calculation]
    C --> D[RESP Fitting]
    D --> E[Replace Charges in ITP]

    B -. "HF/6-31G* or B3LYP/6-31G*" .-> B
    C -. "Merz-Kollman ESP grid" .-> C
    D -. "antechamber -c resp" .-> D

If Gaussian (g16) is available on the system, PRISM executes the entire workflow automatically. Otherwise, it generates a charge.sh script that can be run on a machine with Gaussian installed, and the resulting ligand_resp.mol2 file can be provided back via --respfile.

Water Models¶

TIP3P (Default)¶

prism protein.pdb ligand.mol2 -o output --water tip3p

Best for: AMBER force fields, fastest
Properties: Reasonable density, fast diffusion
Note: Default for most applications

TIP4P¶

prism protein.pdb ligand.mol2 -o output --water tip4p

Best for: OPLS force fields, better properties
Properties: More accurate density, structure
Note: Slightly slower than TIP3P

TIPS3P¶

prism protein.pdb ligand.mol2 -o output --water tips3p

Best for: CHARMM force fields
Properties: Modified TIP3P for CHARMM compatibility
Note: Use with CHARMM36

SPC/E¶

prism protein.pdb ligand.mol2 -o output --water spce

Best for: GROMOS, better diffusion
Properties: Best diffusion constant, good density

Force Field Selection Strategy¶

Decision Tree¶

Start
  |
  ├─ Need highest accuracy?
  |   └─ YES → OpenFF + AMBER14SB
  |   └─ NO → Continue
  |
  ├─ Need fast screening?
  |   └─ YES → GAFF2 + AMBER99SB
  |   └─ NO → Continue
  |
  ├─ Using CHARMM ecosystem?
  |   └─ YES → CGenFF + CHARMM36
  |   └─ NO → Continue
  |
  ├─ Small molecule focus?
  |   └─ YES → OPLS-AA (both)
  |   └─ NO → GAFF2 + AMBER14SB

By Application¶

Binding Free Energy (FEP/TI):

prism protein.pdb ligand.sdf -o output \
  --forcefield amber14sb \
  --ligand-forcefield openff

Virtual Screening (High-throughput):

prism protein.pdb ligand.mol2 -o output \
  --forcefield amber99sb \
  --ligand-forcefield gaff2 \
  --production-ns 100

Membrane Proteins:

prism protein.pdb ligand.mol2 -o output \
  --forcefield charmm36 \
  --ligand-forcefield cgenff \
  --forcefield-path /path/to/cgenff

Novel Chemistry:

# Try in order:
1. OpenFF (best coverage)
2. OPLS-AA (via LigParGen)
3. MMFF/MATCH (SwissParam)

Force Field Compatibility¶

Compatible Combinations¶

Protein FF	Compatible Ligand FFs	Compatible Water
AMBER99SB	GAFF, GAFF2, OpenFF, OPLS	TIP3P, TIP4P, SPC/E
AMBER14SB	GAFF, GAFF2, OpenFF, OPLS	TIP3P, TIP4P
CHARMM36	CGenFF (preferred), GAFF	TIPS3P
OPLS-AA	OPLS-AA (preferred), GAFF	TIP4P, TIP3P

Mixing Force Fields¶

Generally OK: - AMBER protein + GAFF/GAFF2/OpenFF ligand ✓ - CHARMM36 + CGenFF ✓ - OPLS-AA protein + OPLS-AA ligand ✓

Not Recommended: - CHARMM36 + OpenFF ⚠ (different parametrization philosophy) - OPLS-AA + GAFF ⚠ (can work but less validated)

Advanced Topics¶

Custom Force Fields¶

PRISM can install additional force fields:

# Install custom force fields from PRISM configs
prism --install-forcefields

# Interactive installation
# Copies force fields to GROMACS directory

Validating Force Field Choice¶

After building, validate your choice:

import mdtraj as md
import numpy as np

# Load system
traj = md.load("output/GMX_PROLIG_MD/solv_ions.gro")

# Check for clashes (atoms < 0.15 nm apart)
pairs = traj.topology.select_pairs("protein", "resname LIG")
distances = md.compute_distances(traj, pairs)[0]
clashes = np.sum(distances < 0.15)
print(f"Close contacts: {clashes}")

# Check ligand geometry
lig_atoms = traj.topology.select("resname LIG")
lig_traj = traj.atom_slice(lig_atoms)
rgyr = md.compute_rg(lig_traj)[0]
print(f"Ligand radius of gyration: {rgyr:.3f} nm")

Force Field Citation¶

Always cite the force fields you use:

AMBER14SB:

Maier, J.A., Martinez, C., Kasavajhala, K., Wickstrom, L., Hauser, K.E., and Simmerling, C.
(2015). ff14SB: Improving the Accuracy of Protein Side Chain and Backbone Parameters from ff99SB.
J. Chem. Theory Comput. 11, 3696–3713.

OpenFF:

Boothroyd, S., Behara, P.K., Madin, O.C., et al. (2023).
Development and Benchmarking of Open Force Field 2.0.0: The Sage Small Molecule Force Field.
J. Chem. Theory Comput. 19, 3251–3275.

GAFF:

Wang, J., Wolf, R.M., Caldwell, J.W., Kollman, P.A., and Case, D.A. (2004).
Development and testing of a general AMBER force field.
J. Comput. Chem. 25, 1157–1174.

Troubleshooting¶

"Force field not found"¶

# List available force fields
prism --list-forcefields

# Check GROMACS installation
gmx pdb2gmx -h | grep "Force fields"

"Ligand parameterization failed"¶

Try alternatives in order:

# 1. Try OpenFF
prism protein.pdb ligand.sdf -o output --ligand-forcefield openff

# 2. Try OPLS-AA
prism protein.pdb ligand.mol2 -o output --ligand-forcefield opls

# 3. Try SwissParam
prism protein.pdb ligand.mol2 -o output --ligand-forcefield mmff

"Unusual atom types"¶

OpenFF has best coverage:

prism protein.pdb ligand.sdf -o output --ligand-forcefield openff

Force Fields Guide¶

Quick Selection Guide¶

Recommended Combinations¶

Quick Examples¶

Protein Force Fields¶

AMBER Family (Recommended)¶

AMBER99SB¶

AMBER99SB-ILDN¶

AMBER14SB ⭐ Recommended¶

AMBER03¶

CHARMM Family¶

CHARMM36¶

CHARMM27¶

OPLS Family¶

OPLS-AA/L¶

GROMOS Family¶

Ligand Force Fields (8+ Options)¶

Comparison Table¶

1. GAFF (General AMBER Force Field)¶

2. GAFF2 (Improved GAFF)¶

3. OpenFF (Open Force Field) ⭐¶

4. CGenFF (CHARMM General Force Field)¶

5. OPLS-AA (via LigParGen)¶

6-8. SwissParam Force Fields (MMFF, MATCH, Hybrid)¶

Gaussian RESP Charges¶

Theory and Workflow¶

Step 1: Geometry Optimization (Optional)¶

Step 2: Electrostatic Potential (ESP) Calculation¶

Step 3: RESP Charge Fitting¶

Complete Workflow¶

Water Models¶

TIP3P (Default)¶

TIP4P¶

TIPS3P¶

SPC/E¶

Force Field Selection Strategy¶

Decision Tree¶

By Application¶

Force Field Compatibility¶

Compatible Combinations¶

Mixing Force Fields¶

Advanced Topics¶

Custom Force Fields¶

Validating Force Field Choice¶

Force Field Citation¶

Troubleshooting¶

"Force field not found"¶

"Ligand parameterization failed"¶

"Unusual atom types"¶

What's Next¶