AI-Driven Drug Discovery: Target Identification, Virtual Screening, Molecular Docking & ADMET Prediction

Aim

Workshop Objectives

Workshop Structure

📅 Day 1: Target Discovery, Validation and Compound Shortlisting

Focus: Identifying a disease-relevant therapeutic target and building a focused compound library for downstream computational screening.

Topics Covered

• Introduction to AI-assisted target discovery and early-stage drug discovery workflows.
• Disease-gene association and therapeutic target identification.
• Biological relevance, druggability and target-prioritisation criteria.
• Protein-function annotation and pathway-level interpretation.
• Protein-protein interaction network analysis.
• Exploration of chemical and bioactivity databases.
• Retrieval of known ligands, inhibitors and related bioactive compounds.
• Molecular descriptors and physicochemical property analysis.
• Drug-likeness and lead-likeness based compound filtering.
• Development of a focused compound library for virtual screening.

🛠️ Hands-on Activities

• Identify disease-associated genes using public biomedical databases.
• Prioritise a therapeutic protein target based on biological and network evidence.
• Analyse the selected target’s protein interaction network.
• Retrieve relevant bioactive compounds from PubChem and ChEMBL.
• Generate molecular descriptors using RDKit.
• Apply basic physicochemical and drug-likeness filters.
• Build a focused compound shortlist for molecular docking.

🧰 Tools Covered:

GeneCards, UniProt, STRING, PubChem, ChEMBL, RDKit, Python, Google Colab.

📅 Day 2: Virtual Screening, Molecular Docking and Interaction Analysis

Focus: Evaluating shortlisted compounds against the selected protein target using structure-based computational screening and protein-ligand interaction analysis.

Topics Covered

• Principles of structure-based virtual screening and molecular docking.
• Retrieval and evaluation of experimentally determined protein structures.
• Protein cleaning, hydrogen addition and structural preparation.
• Ligand preparation, protonation and geometry optimisation.
• Identification of active sites and ligand-binding pockets.
• Selection of docking-grid coordinates and search-space parameters.
• Configuration of docking exhaustiveness and pose-generation settings.
• Interpretation of docking scores, binding affinities and pose orientations.
• Analysis of hydrogen bonds, hydrophobic interactions, salt bridges and aromatic interactions.
• Comparison of candidate binding modes and interaction profiles.
• Limitations of docking scores and importance of biological interpretation.

🛠️ Hands-on Activities

• Retrieve and prepare the selected protein structure.
• Prepare the shortlisted compounds for docking.
• Identify the binding pocket and define the docking search space.
• Perform molecular docking using AutoDock Vina.
• Compare binding scores and top-ranked poses.
• Visualise protein-ligand interactions.
• Generate an interaction-based ranking of the strongest candidates.

🧰 Tools Covered:

RCSB Protein Data Bank, PubChem, AutoDock Vina, PyRx, PyMOL, Protein-Ligand Interaction Profiler, Google Colab.

📅 Day 3: AI-Based Activity Prediction, ADMET Profiling and Lead Prioritisation

Focus: Integrating machine-learning predictions, molecular features, docking results and ADMET properties to identify the most promising lead candidates.

Topics Covered

• Molecular fingerprints and descriptor-based compound representation.
• Introduction to QSAR and machine-learning-based activity prediction.
• Preparation of molecular datasets for predictive modelling.
• Random Forest and XGBoost algorithms for compound activity prediction.
• Model training, validation and performance evaluation.
• Interpretation of molecular features using SHAP.
• Prediction of drug-likeness and pharmacokinetic properties.
• Assessment of absorption, distribution, metabolism and excretion.
• Toxicity prediction and medicinal chemistry alerts.
• Integration of docking, predicted activity and ADMET results.
• Multi-parameter lead scoring and compound prioritisation.
• Selection of primary and backup lead candidates.

🛠️ Hands-on Activities

• Generate molecular fingerprints and descriptors using RDKit.
• Build and evaluate a machine-learning activity-prediction model.
• Predict biological activity for shortlisted compounds.
• Interpret important molecular features using SHAP.
• Evaluate drug-likeness and ADMET properties using SwissADME.
• Compare docking, predicted activity and pharmacokinetic results.
• Develop an integrated lead-prioritisation matrix.
• Select the final lead and backup candidates.

🧰 Tools Covered:

Python, RDKit, scikit-learn, XGBoost, SHAP, SwissADME, PubChem, Google Colab, Microsoft Excel or Google Sheets.

Who Should Enrol?

Important Dates

Workshop Outcomes