Molecular Docking

Screen and dock small molecules to identify binding modes and affinities

Molecular docking predicts how small molecules bind to protein targets. LiteFold provides powerful docking tools for virtual screening, lead optimization, and binding mode analysis.

Overview

Use molecular docking to:

• Screen large compound libraries
• Identify binding modes and key interactions
• Rank compounds by predicted affinity
• Optimize lead compounds
• Predict selectivity across protein families

Docking Methods

Rigid Docking

Fast screening for large libraries. Protein treated as rigid.

Flexible Docking

Accounts for protein and ligand flexibility. More accurate but slower.

Ensemble Docking

Dock against multiple protein conformations. Best for flexible targets.

Covalent Docking

Model covalent bond formation between ligand and protein.

Quick Start: Docking a Single Compound

1. Prepare Your Protein

   Upload a PDB file or use a structure prediction from LiteFold.

2. Define Binding Site

   Choose one of three methods:

   • Auto-detect: Let Rosalind find binding pockets
   • Residue selection: Specify key binding residues
   • Grid box: Define xyz coordinates and box size
3. Add Your Ligand

   Provide compound as:

   • SMILES string
   • SDF/MOL2 file
   • Draw using molecule sketcher
4. Select Scoring Function

   • AutoDock Vina: Fast, generally reliable
   • Gnina: ML-based, more accurate for some targets
   • Consensus: Combines multiple scoring functions
5. Run Docking

   Click “Dock” and wait for results (typically 1-5 minutes per compound).

Virtual Screening

Screen thousands to millions of compounds in parallel.

Compound Libraries

LiteFold provides access to:

• FDA-approved drugs (~3,000 compounds)
• Enamine REAL (>20 billion compounds)
• ZINC15 (>1 billion compounds)
• ChEMBL (>2 million bioactive compounds)
• Custom libraries: Upload your own

Screening Workflow

1. Select Library

   Choose a pre-built library or upload custom compounds.

2. Apply Filters

   Filter by molecular properties:

   • Molecular weight
   • LogP
   • TPSA
   • Rotatable bonds
   • Drug-likeness (Lipinski, Veber)
3. Configure Docking

   • Binding site definition
   • Scoring function
   • Number of poses per compound
   • Energy cutoffs
4. Launch Screening

   LiteFold parallelizes across cloud compute. Screen 100,000 compounds in hours.

5. Analyze Hits

   • Rank by docking score
   • Filter by interactions
   • Cluster by scaffold
   • Export top hits

Understanding Docking Results

Scoring and Ranking

Docking scores approximate binding affinity. Lower (more negative) is better.

Typical score ranges:

• < -10 kcal/mol: Strong binder
• -8 to -10: Moderate binder
• -6 to -8: Weak binder
• > -6: Likely non-binder

Important: Docking scores are estimates! Always validate top hits experimentally. False positives are common.

Binding Pose Analysis

For each compound, examine:

• Orientation: How ligand sits in pocket
• Hydrogen bonds: Key polar interactions
• Hydrophobic contacts: Non-polar interactions
• π-π stacking: Aromatic interactions
• Salt bridges: Charged interactions

Interaction Maps

LiteFold automatically generates 2D interaction diagrams showing:

• Residues within 4Å of ligand
• Hydrogen bond donors/acceptors
• Hydrophobic contacts
• Metal coordination
• π interactions

Advanced Features

Ensemble Docking

Dock against multiple protein conformations:

1. Generate ensemble from MD simulation or prediction
2. Select representative conformations (5-10 typical)
3. LiteFold docks ligand to all conformations
4. Reports best pose and average score

Benefits:

• Captures induced fit effects
• More accurate for flexible proteins
• Reduces false negatives

Covalent Docking

Model covalent inhibitors:

1. Specify reactive residue (e.g., Cys797 in EGFR)
2. Define warhead chemistry (e.g., acrylamide, chloroacetamide)
3. LiteFold models covalent bond formation
4. Scores combine non-covalent and covalent contributions

Water-Mediated Interactions

Include explicit water molecules:

• Identify conserved waters in binding site
• Include during docking
• Capture water-mediated hydrogen bonds

Cofactor Handling

Automatically handles:

• Metal ions (Mg²⁺, Zn²⁺, Ca²⁺)
• Cofactors (ATP, NAD, heme)
• Post-translational modifications

Customization Options

Receptor Preparation

• Protonation state: Auto or manual pH settings
• Tautomers: Consider alternative forms
• Flexible residues: Select sidechains to move
• Water removal: Keep/remove crystallographic waters

Ligand Preparation

• Conformer generation: Number of starting conformers
• Ionization: Set pH for protonation
• Tautomers: Generate tautomeric forms
• Stereoisomers: Expand undefined stereocenters

Docking Grid

• Center: xyz coordinates or residue-based
• Size: Box dimensions (Å)
• Resolution: Grid spacing (0.3-0.5 Å typical)
• Padding: Extra space around ligand

Best Practices

Use known actives as controls: Always dock a few known binders as positive controls to validate your setup.

Consensus scoring: For important compounds, use multiple scoring functions and look for agreement.

Visual inspection: Don’t rely solely on scores. Inspect binding modes of top hits—do they make chemical sense?

Validate with MD: Run short MD simulations on top hits to check binding stability.

Common pitfalls:

• Oversized binding site: Reduces selectivity
• Wrong protonation state: Misses key interactions
• Ignoring flexibility: Leads to false negatives
• Over-reliance on scores: Must validate experimentally

Scaffold Hopping

Find chemically distinct scaffolds with similar binding:

1. Dock your reference compound
2. Click “Find alternative scaffolds”
3. LiteFold searches for:
   • Different core structures
   • Similar binding mode
   • Comparable predicted affinity
4. Presents diverse hits for expansion

Selectivity Profiling

Dock compounds against multiple protein targets:

• Assess selectivity vs. off-targets
• Identify promiscuous binders
• Predict side effects
• Optimize for selectivity

Example: Dock your kinase inhibitor against a panel of 50 kinases to predict selectivity profile.

ADMET Prediction

For top docking hits, predict ADMET properties:

• Absorption: Caco-2, MDCK permeability
• Distribution: Plasma protein binding, Vd
• Metabolism: CYP inhibition, metabolic stability
• Excretion: Clearance routes
• Toxicity: hERG, AMES, hepatotoxicity

Filter compounds early to avoid downstream failures.

Example: Virtual Screening Campaign

Let’s screen 100,000 compounds against EGFR kinase:

1. Prepare Protein

   • Use AlphaFold prediction or PDB 1M17
   • Define ATP binding site
   • Keep key water molecule (HOH 1077)
2. Select Library

   • Enamine HTS library
   • Filter MW 250-500, LogP < 5
   • 100,000 compounds pass filters
3. Run Screening

   • AutoDock Vina scoring
   • 10 poses per compound
   • Parallel processing: completes in 4 hours
4. Analyze Results

   • 842 compounds score < -8 kcal/mol
   • Cluster into 23 chemical series
   • Visual inspection of top 100
   • 47 compounds selected for synthesis
5. Experimental Validation

   • Purchase or synthesize 47 compounds
   • Test in biochemical assay
   • 8 compounds active (17% hit rate)
   • 3 validated in cell assay

Integration with Other Tools

Docking results flow to:

• Molecular Dynamics: Validate binding with MD
• De Novo Design: Use binding mode as template
• SAR Analysis: Guide medicinal chemistry
• Rosalind AI: Ask “Why does compound A bind better?”

Next Steps

Molecular Dynamics

Validate docking with MD simulations

De Novo Design

Design novel molecules based on docking insights

Binding Affinity

Calculate precise binding free energies

Compound Screening

Complete virtual screening workflows