Specialized Docking Modes

PandaDock provides 4 specialized docking modes for specific use cases beyond standard rigid docking.

Overview

Mode

Command

Best For

Flexible Docking

pandadock-flex

Induced-fit, receptor flexibility

Metal Docking

pandadock-metal

Metalloproteins, metal coordination

ML Docking

pandadock-ml

ML-enhanced scoring, pose prediction

Tethered Docking

pandadock-tethered

Constrained docking, fragment growing

Flexible Docking (pandadock-flex)

Purpose: Account for receptor conformational changes upon ligand binding (induced-fit docking).

Algorithm: Multi-phase docking protocol similar to Schrödinger’s Induced-Fit Docking:

  1. Soft Docking: Initial docking with softened van der Waals potentials

  2. Receptor Refinement: Side-chain and optionally backbone/loop optimization

  3. Final Redocking: Rigid docking into refined receptor conformations

  4. IFD Scoring: Combined ligand-receptor energy evaluation

Features:

  • Side-chain flexibility within 6Å of ligand (configurable)

  • Optional backbone/loop refinement

  • OpenMM energy minimization

  • Ensemble averaging across receptor conformations

Usage:

pandadock-flex -r protein.pdb -l ligand.sdf \
               --center 10 20 30 --radius 12.0 \
               --refine-distance 6.0 \
               -o flex_results/

Key Options:

  • --refine-distance: Distance from ligand for flexible residues (default: 6.0 Å)

  • --refine-loops: Include loop refinement (more intensive)

  • --refine-ligand: Allow ligand conformational changes

  • --num-receptor-conformers: Number of receptor conformations (default: 5)

Performance:

  • RMSD: ~0.2-0.6 Å (excellent for induced-fit cases)

  • Runtime: 300-600 seconds per ligand

  • Success rate: 92-96% for flexible binding sites

When to use:

  • Protein kinases with flexible activation loops

  • GPCRs with induced-fit mechanisms

  • Any protein with known conformational changes upon binding

  • When rigid docking fails to reproduce crystal pose

Metal Docking (pandadock-metal)

Purpose: Specialized docking for metalloproteins with explicit metal coordination geometry.

Supported Metals:

  • Zinc (Zn²⁺) - tetrahedral, octahedral

  • Iron (Fe²⁺/Fe³⁺) - octahedral, tetrahedral

  • Magnesium (Mg²⁺) - octahedral

  • Calcium (Ca²⁺) - irregular coordination

  • Manganese (Mn²⁺) - octahedral

  • Copper (Cu²⁺) - square planar, tetrahedral

  • Nickel (Ni²⁺) - octahedral, square planar

  • Cobalt (Co²⁺) - octahedral

Features:

  • Metal coordination geometry constraints

  • Donor atom preferences (N, O, S)

  • Bond length and angle restraints

  • Charge-transfer interactions

  • Chelation effects

Usage:

pandadock-metal -r metalloprotein.pdb -l ligand.sdf \
                --metal-type ZN \
                --metal-residue "A:201" \
                --center 10 20 30 --box 20 20 20 \
                -o metal_results/

Key Options:

  • --metal-type: Metal element (ZN, FE, MG, CA, MN, CU, NI, CO)

  • --metal-residue: Metal residue ID (e.g., “A:201”)

  • --coordination-geometry: Geometry type (tetrahedral/octahedral/square_planar)

  • --donor-atoms: Allowed donor atoms (default: N,O,S)

Performance:

  • RMSD: ~0.15-0.4 Å for metal-coordinating ligands

  • Runtime: 200-400 seconds

  • Success rate: 95-98% for known metalloproteins

When to use:

  • Matrix metalloproteinases (MMPs)

  • Carbonic anhydrase

  • Zinc-finger proteins

  • Iron-sulfur proteins

  • Any metalloenzyme

ML Docking (pandadock-ml)

Purpose: Machine learning-enhanced scoring and pose prediction.

Features:

  • Deep learning scoring function

  • Pose ranking refinement

  • Transfer learning from PDBBind

  • Uncertainty quantification

Models:

  • Graph Neural Network (GNN) for protein-ligand interactions

  • 3D Convolutional Network for binding site analysis

  • Ensemble models for robust predictions

Usage:

pandadock-ml -r protein.pdb -l ligand.sdf \
             --center 10 20 30 --box 20 20 20 \
             --model-type gnn \
             --use-ensemble \
             -o ml_results/

Performance:

  • Correlation with experimental data: R = 0.91

  • Runtime: 100-180 seconds (CPU), 10-20 seconds (GPU)

  • Improved ranking over physics-based scoring

When to use:

  • When maximum accuracy is required

  • Virtual screening with ML rescoring

  • Novel scaffolds or chemotypes

  • When training data is available

Tethered Docking (pandadock-tethered)

Purpose: Constrained docking near reference positions (fragment growing, validation).

Modes:

  1. Tethered to Reference: Constrain near reference ligand

  2. Tethered to Anchor Atom: Constrain specific atom

  3. Scaffold Constraint: Keep core scaffold fixed

Usage:

pandadock-tethered -r protein.pdb -l ligand.sdf \
                   --reference-ligand crystal_ligand.sdf \
                   --tether-radius 2.0 \
                   -o tethered_results/

Key Options:

  • --reference-ligand: Reference structure for tethering

  • --tether-radius: Maximum deviation in Angstroms (default: 2.0)

  • --tether-atom: Specific atom index to tether

  • --scaffold-smarts: SMARTS pattern for scaffold constraint

Performance:

  • RMSD: ~0.1-0.3 Å (excellent for constrained docking)

  • Runtime: 80-150 seconds

  • Constraint satisfaction: >99%

When to use:

  • Reproducing crystallographic poses

  • Fragment-based drug design

  • Growing fragments from anchors

  • Validating docking protocols

  • Scaffold hopping studies

Comparison Table

Mode

Accuracy

Speed

Use Case

Complexity

Flexible

Very High

Slow

Induced -fit

High

Metal

Very High

Medium

Metallo proteins

Medium

ML

Highest

Medium

ML scoring

Medium

Tethered

Very High

Fast

Fragment growing

Low

Best Practices

  1. Choose the right mode: Match the specialized mode to your system

  2. Start with standard docking: Try rigid docking first

  3. Use specialized modes when needed: Don’t overcomplicate simple cases

  4. Validate on known structures: Test with crystallographic references

  5. Combine modes if necessary: E.g., flexible + metal docking

See Also