pandadock-flex - Flexible Docking Command

The pandadock-flex command performs induced-fit docking with receptor flexibility. It accounts for protein conformational changes upon ligand binding, similar to Schr?dinger’s Induced-Fit Docking (IFD).

Synopsis

pandadock-flex [OPTIONS]

Description

Performs flexible docking using a multi-phase protocol:

Soft Docking: Initial docking with softened van der Waals potentials
Receptor Refinement: Side-chain and optional backbone optimization
Final Redocking: Rigid docking into refined receptor conformations
IFD Scoring: Combined ligand-receptor energy evaluation

This approach is essential for targets with induced-fit binding mechanisms where the receptor undergoes conformational changes upon ligand binding.

Required Options

-r, --receptor PATH: Receptor PDB file (protein structure)
-l, --ligand PATH: Ligand file (SDF, MOL2, or PDB format)
--center X Y Z: Grid box center coordinates (X Y Z in Angstroms)
--radius FLOAT: Grid box radius in Angstroms. Default: 12.0

Flexibility Options

--refine-distance FLOAT

Distance from ligand for flexible residues (Angstroms). Default: 6.0

Residues within this distance will have side-chain flexibility.

--refine-loops / --no-refine-loops

Include loop refinement. Default: disabled

Enable for targets with flexible loops near binding site.

--refine-backbone / --no-refine-backbone

Include backbone refinement. Default: disabled

WARNING: Very computationally intensive. Use sparingly.

--refine-ligand / --no-refine-ligand

Allow ligand conformational changes during refinement. Default: enabled

--num-receptor-conformers N

Number of receptor conformations to generate. Default: 5

Higher values increase accuracy but also runtime.

--flexible-residues RESIDUES

Explicitly specify flexible residues (e.g., “A:100,A:150,B:200”)

Overrides automatic distance-based selection.

Docking Algorithm

-a, --algorithm ALGORITHM

Docking algorithm for initial and final docking. Default: enhanced_hierarchical_cpu

Recommended: enhanced_hierarchical_cpu, genetic_algorithm_cpu

Scoring Options

-s, --scoring FUNCTION

Scoring function. Default: physics_based

Options: physics_based, hybrid, precision_score

--ifd-scoring / --no-ifd-scoring

Use Induced-Fit Docking scoring (ligand + receptor energy). Default: enabled

Output Options

-o, --output-dir PATH: Output directory for results. Default: flex_docking_output
-n, --num-poses N: Number of final poses to generate. Default: 20
--visualize / --no-visualize: Generate visualization plots. Default: enabled
--save-receptor-conformers: Save all refined receptor conformations

Performance Options

--cpuworkers N: Number of CPU worker threads. Default: auto-detect
--gpu: Enable GPU acceleration for docking steps (requires CUDA)
--fast: Fast mode with reduced sampling (for testing)

Examples

Basic Flexible Docking

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

This uses default settings:

6? side-chain flexibility
5 receptor conformers
OpenMM refinement

Kinase Flexible Docking

pandadock-flex -r kinase.pdb -l inhibitor.sdf \\
               --center 15 20 25 --radius 12.0 \\
               --refine-distance 8.0 \\
               --refine-loops \\
               --num-receptor-conformers 10 \\
               -o kinase_flex/

Includes loop refinement (important for DFG-in/out transitions)

GPCR Flexible Docking

pandadock-flex -r gpcr.pdb -l agonist.sdf \\
               --center 12 15 18 --radius 14.0 \\
               --refine-distance 7.0 \\
               --refine-backbone \\
               --num-receptor-conformers 15 \\
               --refinement-steps 2000 \\
               -o gpcr_flex/

Extended refinement for GPCR conformational changes

High-Accuracy Flexible Docking

pandadock-flex -r protein.pdb -l ligand.sdf \\
               --center 10 20 30 --radius 12.0 \\
               --algorithm enhanced_hierarchical_cpu \\
               --scoring hybrid \\
               --refine-distance 8.0 \\
               --num-receptor-conformers 20 \\
               --refinement-steps 2000 \\
               --num-poses 50 \\
               -o high_accuracy_flex/

Fast Testing Mode

pandadock-flex -r protein.pdb -l ligand.sdf \\
               --center 10 20 30 --radius 12.0 \\
               --fast \\
               --num-receptor-conformers 3 \\
               --refinement-steps 500 \\
               -o fast_test/

GPU-Accelerated Flexible Docking

pandadock-flex -r protein.pdb -l ligand.sdf \\
               --center 10 20 30 --radius 12.0 \\
               --algorithm enhanced_hierarchical_gpu \\
               --gpu \\
               -o gpu_flex/

Custom Flexible Residues

pandadock-flex -r protein.pdb -l ligand.sdf \\
               --center 10 20 30 --radius 12.0 \\
               --flexible-residues "A:99,A:145,A:189,B:201" \\
               -o custom_flex/

Output Files

The command generates the following outputs:

Structures:

complex1.pdb, complex2.pdb, ... - Protein-ligand complexes with refined receptor
pose1.pdb, pose2.pdb, ... - Ligand poses only
receptor_refined_*.pdb - Refined receptor conformations (if --save-receptor-conformers)

Analysis:

flex_docking_results.json - Complete results with IFD scores
refinement_analysis.json - Receptor refinement details
ifd_scores.csv - IFD scoring breakdown
summary.txt - Human-readable summary

Visualizations:

ifd_scores.png - IFD score distribution
receptor_rmsd.png - Receptor conformational changes
interaction_map.png - Protein-ligand interaction analysis

Performance Characteristics

Runtime:

Small ligand, 5 conformers: 5-10 minutes
Medium ligand, 10 conformers: 15-30 minutes
Large ligand, 20 conformers: 30-60 minutes

With loop refinement: 2-3x longer

With backbone refinement: 5-10x longer

Accuracy:

RMSD: 0.2-0.6 ? (excellent for induced-fit cases)
Success rate: 92-96% for flexible binding sites

Best Practices

When to Use Flexible Docking

Protein kinases with flexible activation loops

GPCRs with induced-fit activation

Proteins with known conformational changes upon binding

When rigid docking fails to reproduce crystal pose (RMSD > 2?)

Large, conformationally diverse ligands (peptides, macrocycles)

When Not to Use

Small, rigid binding sites - Use standard rigid docking

Virtual screening - Too slow for large libraries

Well-defined binding modes - Rigid docking sufficient

Optimization Tips

Balance accuracy and speed:

# Fast but reasonable
--num-receptor-conformers 5 --refinement-steps 1000

# High accuracy
--num-receptor-conformers 15 --refinement-steps 2000

# Ultra-high accuracy
--num-receptor-conformers 20 --refinement-steps 3000

Refine only necessary regions:

# Auto-select within 6? (default)
--refine-distance 6.0

# Larger binding site
--refine-distance 8.0

Use GPU when available:

--algorithm enhanced_hierarchical_gpu --gpu

Troubleshooting

Slow Performance

Problem: Flexible docking takes too long

Solutions:

Reduce receptor conformers: --num-receptor-conformers 3
Reduce refinement steps: --refinement-steps 500
Use --fast mode
Disable loop refinement
Use GPU: --gpu

Poor Results

Problem: Poor pose prediction or binding affinity