rdk-amber

Small utilities for bridging AMBER force-field conventions and RDKit:

1. A mapping between AMBER / GAFF atom types and chemical elements.
2. A loader that reads an AMBER-style mol2 file (GAFF/AMBER atom types in the type column) into an rdkit.Chem.Mol.

RDKit's own MolFromMol2File expects SYBYL atom types (C.3, N.ar, …). AmberTools (antechamber, tleap) instead writes GAFF/AMBER atom types (c3, na, …) in that column, which RDKit cannot interpret. This package parses such files directly.

Install

pip install -e .

Requires rdkit (install via conda/pip).

Usage

Atom type → element

from rdk_amber import element_from_amber_type

element_from_amber_type("c3")    # 'C'
element_from_amber_type("ho")    # 'H'
element_from_amber_type("Na+")   # 'Na'
element_from_amber_type("Xx")    # None  (unknown)
element_from_amber_type("Xx", strict=True)   # raises UnknownAtomTypeError

The full table is exposed as rdk_amber.AMBER_ATOM_TYPE_TO_ELEMENT. Types not in the table fall back to a prefix heuristic (e.g. c99 → C).

Load an AMBER mol2

from rdk_amber import load_amber_mol2

mol = load_amber_mol2("ligand.mol2")          # sanitized RDKit Mol, with 3D coords
mol = load_amber_mol2("ligand.mol2", remove_hs=True)

Or use the class for more control / multi-molecule files:

from rdk_amber import AmberMol2Loader

loader = AmberMol2Loader.from_file("ligand.mol2")
loader.num_molecules        # number of @<TRIPOS>MOLECULE records
loader.names                # their names
mol  = loader.to_rdkit(index=0, sanitize=True)
mols = loader.to_rdkit_all()

Each atom keeps the original AMBER data as properties:

and the molecule records the charge state that was reconstructed:

Formal charges

The Tripos mol2 format has no formal-charge column, so a protonated amine and a neutral one have identical connection tables. Left alone, RDKit rejects the cation (Explicit valence for atom # N, 4, is greater than permitted) and — the quieter problem — reads a carboxylate back as the neutral acid, because a singly-bonded O is a perfectly legal hydroxyl that it fills with an implicit H.

What the format does record is a partial charge per atom, and for a force-field-ready structure those sum to the net charge. That sum is the authority for how much charge to place; the sites come from RDKit's own valence model (over-valent atoms) and from the terminal O/S of the usual acidic groups. Sums are taken per connected fragment, so a salt charges each component separately, and over-valent atoms are charged before anions are balanced against the net, so zwitterions come out with both signs.

mol = load_amber_mol2("ammonium.mol2")
mol.GetIntProp("_AmberNetCharge")            # +1

Pass assign_charges=False for the previous behaviour. If a file's bond orders and partial charges disagree about the protonation state, the molecule is still built, an AmberChargeWarning is raised, and the two net-charge properties differ — compare them to reject such a structure without catching warnings.

Notes

• Bond orders are read from the @<TRIPOS>BOND section. ar → aromatic, am → single, 1/2/3 → single/double/triple.
• If sanitization fails (unusual force-field valences), pass sanitize=False to obtain the unsanitized molecule.

Tests

pytest

rdk-amber