AI-Powered Drug Discovery: Molecular Screening, ADMET Prediction, QSAR Modeling & Lead Optimization

Aim

Workshop Objectives

Workshop Structure

📅 Day 1: AI-Driven Molecular Screening and Compound Prioritization

• Drug discovery pipeline: target identification to lead optimization
• Role of AI and machine learning in early-stage drug discovery
• Molecular representations: SMILES, molecular descriptors, and fingerprints
• Drug-likeness assessment using physicochemical properties
• Lipinski’s Rule of Five and compound filtering
• Virtual screening and molecular-similarity analysis
• Ranking compounds for further investigation

🛠️ Hands-on: Build a molecular-screening pipeline in Google Colab to retrieve compounds, calculate molecular descriptors, apply drug-likeness filters, and rank candidate molecules.

🧰 Tools Covered: Google Colab, Python, RDKit, Pandas, and PubChem PUG REST API

📅 Day 2: Machine Learning for ADMET Prediction

• Understanding absorption, distribution, metabolism, excretion, and toxicity
• Importance of ADMET evaluation in reducing late-stage drug failure
• Preparing molecular datasets for predictive modelling
• Descriptor and fingerprint-based feature engineering
• QSAR and machine-learning approaches for ADMET prediction
• Model training using Random Forest and classification algorithms
• Performance evaluation using ROC-AUC, accuracy, precision, and recall
• Interpreting ADMET predictions and identifying risk alerts

🛠️ Hands-on: Train and evaluate a machine-learning model in Google Colab to predict an ADMET-related property and classify molecules as favourable or high-risk candidates.

🧰 Tools Covered: Google Colab, RDKit, Scikit-learn, Pandas, and Matplotlib

📅 Day 3: Lead Prediction, Explainable AI and Multi-Parameter Optimization

• Lead identification and compound-ranking strategies
• Integrating activity, drug-likeness, and ADMET predictions
• Multi-parameter optimization for lead selection
• AI-based scoring and ranking of candidate compounds
• Explainable AI for understanding molecular predictions
• Introduction to graph neural networks and generative AI in drug design
• Model limitations, data quality, applicability domains, and validation
• Designing an end-to-end AI-powered drug-discovery workflow

🛠️ Hands-on: Develop a lead-prioritization notebook that combines predicted activity, drug-likeness, and ADMET scores to rank compounds and visualize the strongest candidates.

🧰 Tools Covered: Google Colab, Python, RDKit, Scikit-learn, SHAP, and Seaborn

Important Dates

Workshop Outcomes