Algorithms Overview =================== PandaDock v3.0 provides a streamlined set of docking algorithms optimized for accuracy and reliability. Primary Algorithm ----------------- **Hierarchical Docking (``pandadock``)** The main docking algorithm uses a multi-stage hierarchical search: 1. **Coarse Sampling**: Fast grid-based sampling at multiple resolutions 2. **Refinement**: Progressive refinement of promising regions 3. **Scoring**: Energy evaluation with Vina-style scoring 4. **Ranking**: Final ranking with detailed scoring This algorithm achieves: * Pearson R = 0.12 on ULVSH benchmark * Best traditional docking performance * Consistent pose generation Usage: .. code-block:: bash pandadock dock -r protein.pdb -l ligand.sdf \ --center 10 20 30 --box 20 20 20 Scoring Functions ----------------- **Vina Scoring (Default)** Implements AutoDock Vina-style scoring with: * Gaussian steric interactions (gauss1, gauss2) * Repulsion for steric clashes * Hydrophobic interactions * Hydrogen bonding * Rotatable bond flexibility penalty .. code-block:: bash pandadock dock -r protein.pdb -l ligand.sdf \ --scoring vina ... **Physics-Based Scoring** Alternative scoring with: * Van der Waals (Lennard-Jones) * Electrostatics (Coulomb) * Hydrogen bonding * Solvation effects .. code-block:: bash pandadock dock -r protein.pdb -l ligand.sdf \ --scoring physics_based ... ML Scoring (Recommended) ------------------------ **PandaDock-GNN** For best accuracy, use the SE(3)-equivariant GNN scoring function: * Pearson R = 0.67 on ULVSH benchmark * **5x better** than traditional scoring * Outperforms all baseline methods Use via hybrid docking: .. code-block:: bash pandadock hybrid -r protein.pdb -l ligand.sdf \ --center 10 20 30 --box 20 20 20 \ -m model.pt See :doc:`../gnn/overview` for details. Specialized Modes ----------------- **Flexible Docking** (``pandadock-flex``) Induced-fit docking with receptor flexibility: .. code-block:: bash pandadock-flex -r protein.pdb -l ligand.sdf \ --center 10 20 30 --radius 15 **Metal Docking** (``pandadock-metal``) Specialized for metalloproteins (Zn, Fe, Mg, Ca, etc.): .. code-block:: bash pandadock-metal dock -r metalloprotein.pdb -l ligand.sdf ... **Tethered Docking** (``pandadock-tethered``) Constrained docking for validation: .. code-block:: bash pandadock-tethered dock -r protein.pdb -l ligand.sdf \ --reference crystal_ligand.pdb Algorithm Selection Guide ------------------------- +---------------------------+-----------------------------------+ | Use Case | Recommended Approach | +===========================+===================================+ | **Best accuracy** | ``pandadock hybrid`` with GNN | +---------------------------+-----------------------------------+ | **Quick docking** | ``pandadock dock --fast`` | +---------------------------+-----------------------------------+ | **Virtual screening** | ``pandadock hybrid`` (batch) | +---------------------------+-----------------------------------+ | **Flexible receptor** | ``pandadock-flex`` | +---------------------------+-----------------------------------+ | **Metal coordination** | ``pandadock-metal`` | +---------------------------+-----------------------------------+ | **Pose validation** | ``pandadock-tethered`` | +---------------------------+-----------------------------------+ Performance Comparison ---------------------- On ULVSH benchmark (942 compounds): +---------------------------+-----------------+ | Method | Pearson R | +===========================+=================+ | **PandaDock-GNN (Hybrid)**| **0.67** ⭐ | +---------------------------+-----------------+ | VM2 | 0.15 | +---------------------------+-----------------+ | PandaDock (Hierarchical) | 0.12 | +---------------------------+-----------------+ | PM6 | 0.08 | +---------------------------+-----------------+ | Gnina, MMPBSA, etc. | < 0.02 | +---------------------------+-----------------+