Bio Pose Validation

Version Compatibility

Reference examples tested with: PoseBusters 0.6+, RDKit 2024.09+, pandas 2.2+, posecheck 0.5+ (optional).

Before using code patterns, verify installed versions match. If versions differ:

• Python: pip show <package> then help(module.function) to check signatures

If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.

Pose Validation

Test docked or AI-generated protein-ligand poses for physical plausibility. PoseBusters (Buttenschoen et al. 2024) provides geometric, chemical, and energetic checks that flag implausible poses, including non-planar aromatic rings, van der Waals clashes, broken bonds, altered stereochemistry, and unfavorable internal energies. On the Astex Diverse Set, DiffDock achieved 72% RMSD success but only 47% combined RMSD-and-PB-valid success; the size of this gap is dataset- and method-dependent. PB-valid status complements RMSD for downstream SAR, FEP setup, or generative-model training.

For docking, see chemoinformatics/virtual-screening. For ML docking specifically, see chemoinformatics/ml-docking-rescoring.

PoseBusters Test Suite

PoseBusters runs ~20 individual checks grouped into:

The thresholds below are the benchmark criteria reported by Buttenschoen et al. (2024). Installed PoseBusters defaults may differ by version and configuration, so record the package version and resolved configuration.

A pose passing ALL tests is "PB-valid". Combined PB-valid + RMSD <= 2 Å is the modern criterion.

When to Apply PoseBusters

PoseBusters Usage

from posebusters import PoseBusters

bust = PoseBusters(config='redock')

results = bust.bust(
    mol_pred='predicted.sdf',
    mol_true='reference.sdf',
    mol_cond='receptor.pdb',
)

Common configurations and their included checks are:

PoseBusters also ships additional and faster configurations in some releases. Treat the table as a workflow guide, not an exhaustive registry, and inspect the configurations available in the installed version.

Output: a DataFrame with one row per pose, metadata columns, and boolean pass/fail columns for the checks enabled by the selected configuration. Reference-dependent fields such as RMSD and the exact check-column names vary by configuration and version; inspect results.columns rather than relying on a fixed exhaustive list.

Python Library API

Goal: Programmatically validate a docked-pose SDF against a receptor PDB and produce a PB-valid filter.

Approach: Instantiate PoseBusters(config='dock'), call bust() on the SDF + PDB pair, and AND-aggregate all boolean check columns into a single pb_valid flag.

from posebusters import PoseBusters
import pandas as pd

bust = PoseBusters(config='dock')

results = bust.bust(
    mol_pred='/path/to/docked_poses.sdf',
    mol_cond='/path/to/receptor.pdb',
)

check_cols = [
    col for col in results.select_dtypes(include='bool').columns
    if not col.lower().startswith('rmsd')
]
results['pb_valid'] = results[check_cols].all(axis=1)
valid = results[results['pb_valid']]
print(f'{len(valid)} / {len(results)} poses are PB-valid')

Strain Energy Quantification

Beyond binary PB-valid, quantitative strain energy distinguishes "marginal" from "egregious" poses.

Goal: Quantify how far each docked pose is from its lowest-energy free conformer in MMFF94 energy units.

Approach: Generate a reference conformer ensemble (ETKDGv3 + MMFF94), make the docked and reference molecules chemically consistent by adding explicit hydrogens to both, relax only the added docked-pose hydrogens while fixing all heavy atoms, take the lowest sampled reference energy as baseline, and report docked_energy - min_ref_energy as a relative strain diagnostic. This is not a rigorous solution-phase conformational free energy.

from rdkit import Chem
from rdkit.Chem import AllChem

def ligand_strain(docked_sdf, n_ref=20):
    suppl = Chem.SDMolSupplier(docked_sdf, removeHs=False)
    strains = []
    for docked in suppl:
        if docked is None:
            continue

        smi = Chem.MolToSmiles(docked)
        ref = Chem.MolFromSmiles(smi)
        if ref is None:
            strains.append({'strain': None, 'note': 'reference_parse_failed'})
            continue
        ref = Chem.AddHs(ref)
        props_ref = AllChem.MMFFGetMoleculeProperties(ref)
        if props_ref is None:
            strains.append({'strain': None, 'note': 'no_reference_mmff_parameters'})
            continue
        conf_ids = list(AllChem.EmbedMultipleConfs(
            ref, numConfs=n_ref, params=AllChem.ETKDGv3()
        ))
        if not conf_ids:
            strains.append({'strain': None, 'note': 'reference_embedding_failed'})
            continue
        AllChem.MMFFOptimizeMoleculeConfs(ref)

        ref_energies = []
        for c in conf_ids:
            ff = AllChem.MMFFGetMoleculeForceField(
                ref, props_ref, confId=c
            )
            if ff is not None:
                ref_energies.append(ff.CalcEnergy())
        if not ref_energies:
            strains.append({'strain': None, 'note': 'reference_force_field_failed'})
            continue
        min_ref = min(ref_energies)

        # MMFF energies are comparable only for the same explicit atom system.
        # Add any missing H coordinates, then relax H atoms while preserving the
        # docked heavy-atom pose.
        docked_h = Chem.AddHs(Chem.Mol(docked), addCoords=True)
        if docked_h.GetNumAtoms() != ref.GetNumAtoms():
            strains.append({'strain': None, 'note': 'atom_system_mismatch'})
            continue
        props_docked = AllChem.MMFFGetMoleculeProperties(docked_h)
        docked_ff = AllChem.MMFFGetMoleculeForceField(
            docked_h, props_docked
        ) if props_docked is not None else None
        if docked_ff is not None:
            for atom in docked_h.GetAtoms():
                if atom.GetAtomicNum() != 1:
                    docked_ff.AddFixedPoint(atom.GetIdx())
            docked_ff.Minimize(maxIts=200)
        docked_e = docked_ff.CalcEnergy() if docked_ff else None

        strains.append({
            'min_ref_energy': min_ref,
            'docked_energy': docked_e,
            'strain': docked_e - min_ref if docked_e is not None else None,
            'note': 'ok' if docked_e is not None else 'docked_force_field_failed',
        })
    return strains

Interpret relative MMFF strain in the context of ligand chemistry, conformer-sampling coverage, and force-field support. Boström et al. (1998) found a conformational energy penalty of no more than 3 kcal/mol for about 70% of 33 protein-bound ligands; that result does not establish a universal acceptance cutoff. Treat unusually high values as a prompt for inspection or use a project-defined threshold validated for the series.

vdW Overlap with Protein

The 2024 benchmark criterion limits protein-ligand overlap to 7.5% of the ligand vdW volume, using protein radii scaled by 0.8. PoseBusters' bust(...) computes this check; do not substitute an unvalidated pairwise-distance sketch for its volume calculation.

Aromatic Ring Planarity

import numpy as np

def aromatic_planarity(mol):
    deviations = []
    for ring in mol.GetRingInfo().AtomRings():
        ring_atoms = [mol.GetAtomWithIdx(i) for i in ring]
        if not all(a.GetIsAromatic() for a in ring_atoms):
            continue
        coords = np.array([mol.GetConformer().GetAtomPosition(i)
                          for i in ring])
        centroid = coords.mean(axis=0)
        centered = coords - centroid
        _, s, vh = np.linalg.svd(centered)
        normal = vh[-1]
        deviation = np.abs(centered @ normal).max()
        deviations.append(deviation)
    return max(deviations) if deviations else 0

Aromatic ring deviation > 0.25 Å is implausible; flag.