Virtual Ligand Screening (VLS) in ICM-Pro

This tutorial outlines the structure-based virtual screening workflow, moving from receptor preparation to multi-million to billion sized compound library execution.

1. Receptor & Project Setup

Virtual screening requires a pre-calculated grid potential map of the binding pocket to ensure high-speed docking.

• Prepare Receptor: Load your protein (e.g., 1m17), convert to an ICM object, and optimize hydrogens/protonation states [00:13:00].
• Initialize VLS Project: Go to Docking > New Project. Define the receptor object and the site (using Pocket Finder or an existing ligand) [00:15:49].
• Generate Maps: Set the grid box (Purple Box) to encompass the full pocket. Click Go to calculate maps for H-bonding, van der Waals, hydrophobic, and electrostatic potentials [00:18:15].

2. Database Preparation & Indexing

For screening efficiency, raw SD files should be converted into an indexed (.inx) format.

• Indexing: Navigate to Docking > Tools > Index Mole/Mol2 File. This allows ICM to access records randomly rather than sequentially [00:23:50].
• VLS Setup: Tell the project which database to use via Docking > Set up Batch Ligands [00:24:50].

3. Screening Preferences & Filters

Apply filters on-the-fly to save disk space and focus on drug-like candidates.

• Score Threshold: Set the VLS score cutoff (default is -32). It is recommended to redock the native ligand first to establish a baseline [00:27:32].
• Lipinski Rules: Use Docking > Preferences > Database Scan to filter by molecular weight, LogP, or H-bond counts [00:29:04].
• Charge Prediction: Set Charge Groups to Auto to use the ICM PKA model for proper protonation at pH 7.4 [00:31:52].

4. Execution: GUI & Command Line

Graphical Interface (GUI)

Best for smaller screens (thousands to 100k compounds). Use Docking > Run Docking VLS Batch. You can split the job across multiple local CPUs [00:36:47].

Command Line for In-house Cluster or Cloud

For millions of compounds or cluster usage (SGE/Slurm):

• doc_scan: Use the _dockScan script for automated screening. Use the proc=N argument to parallelize [00:42:33].
• doc_sub: Use _doc_sub to automatically generate queue submission scripts for high-performance clusters [00:45:53].

5. Hit List Analysis

Once complete, use Docking > Make Hit List to aggregate results into a searchable table [00:50:00].

• VLS Score: The physics-based energy function.
• RT CNN Score: A neural network score trained on protein-ligand geometries. It is often more robust against minor clashes than the physics score [00:54:41].
• Multi-Scoring: The best results often come from the intersection of the top 10-20% of both VLS and RTCNN scores [00:55:54].

6. Hitlist Filtering Tools

Filter your results using 3D pharmacophoric similarity and interaction fingerprints.

• H-Bond Filtering: Pick a specific receptor atom and flag all hits that form a hydrogen bond to it [00:57:48].
• APF Pose Similarity: Compare docked poses to a known crystal ligand using 3D Atomic Property Fields [01:00:10].
• Clustering: Use the APF Cluster tool to group hits by pharmacophoric similarity, allowing you to select a diverse set of compounds for testing [01:01:46].