RDKit

1. Overview

The core workflow is SMILES -> Mol -> AddHs -> ETKDG -> MMFF/UFF -> clustering -> SDF. RDKit uses molecular graphs, empirical rules, and force fields; its force-field ranking is not a substitute for electronic energies.

Identifiers, fingerprints, charges, and SMARTS searching are covered in the cheminformatics track.

Because force-field energies and quantum-chemical electronic energies have different definitions, use RDKit primarily to prepare a diverse starting ensemble. Refine several low-energy, structurally distinct candidates at the electronic-structure level when relative energies matter.

2. SMILES and Mol objects

Branches, ring closures, aromatic atoms, formal charges, and stereochemistry are encoded in SMILES. Parsing normally performs sanitization and assigns valence, aromaticity, conjugation, and ring information.

from rdkit import Chem

mol = Chem.MolFromSmiles("C[C@H](O)C(=O)O")
if mol is None:
    raise ValueError("SMILES parsing failed")
print(Chem.MolToSmiles(mol, isomericSmiles=True))

3. Mol objects and sanitization

An RDKit Mol stores atoms, bonds, properties, and zero or more conformers. Parsing normally sanitizes valence, aromaticity, conjugation, hybridization, and ring information. A failed parse returns None; do not continue with an invalid object.

mol = Chem.MolFromSmiles(smiles)
if mol is None:
    raise ValueError(f"Could not parse: {smiles}")
mol_h = Chem.AddHs(mol)
print(mol_h.GetNumAtoms(), mol_h.GetNumConformers())

4. Two-dimensional depiction

Compute 2D coordinates before drawing and use atom or bond highlighting to communicate a specific substructure. Keep orientation consistent when comparing a molecular series.

from rdkit.Chem import AllChem, Draw

AllChem.Compute2DCoords(mol)
Draw.MolToFile(mol, "molecule.svg", size=(420, 280))

5. ETKDG conformer generation

ETKDG combines distance geometry with experimental torsional preferences. Add explicit hydrogens, generate multiple conformers with a fixed random seed, and inspect embedding failures and stereochemistry.

from rdkit.Chem import AllChem

mol_h = Chem.AddHs(mol)
params = AllChem.ETKDGv3()
params.randomSeed = 2026
ids = AllChem.EmbedMultipleConfs(mol_h, numConfs=100, params=params)

Embedding returns conformer identifiers; compare their count with the requested count and preserve the random seed. Pruning during embedding uses an RMS threshold, so record pruneRmsThresh if it is enabled.

6. MMFF and UFF

MMFF is commonly used for organic molecules when parameters are available; UFF covers a broader element range but still requires chemical validation. Record convergence and retain several low-energy, geometrically distinct conformers.

results = AllChem.MMFFOptimizeMoleculeConfs(
    mol_h, numThreads=0, maxIters=1000, mmffVariant="MMFF94s"
)
# Each tuple is (not_converged, energy).
conformer_data = list(zip(ids, results))

A status value of 0 indicates convergence. Check MMFFHasAllMoleculeParams before MMFF; use UFF only after confirming that its atom types are chemically sensible for the system.

7. RMSD and conformer selection

RMSD depends on atom mapping, alignment, symmetry handling, and whether hydrogens are included. Energy windows and RMSD clustering serve different purposes and are often combined.

from rdkit.ML.Cluster import Butina

rms = AllChem.GetConformerRMSMatrix(mol_h, prealigned=False)
clusters = Butina.ClusterData(
    rms, len(ids), 0.75, isDistData=True, reordering=True
)

8. Complete preparation example

1. Parse and validate the molecular graph.
2. Assign or verify stereochemistry.
3. Add explicit hydrogens.
4. Generate many ETKDG conformers with a fixed seed.
5. Optimize each conformer and record convergence.
6. Apply an energy window and RMSD clustering.
7. Write conformer ID, energy, and settings to SDF properties.

10. References

• Getting Started with the RDKit in Python
• The RDKit Book
• ETKDGv3 paper
• RDKit distance-geometry API
• RDKit force-field helpers