📅 Day 1: AlphaFold 3 Structure Prediction and Target Preparation
Focus: Retrieving, evaluating and preparing reliable protein structures for molecular docking.
- Introduction to AI-powered protein structure prediction and structure-based drug discovery.
- Working principles and applications of AlphaFold 3.
- Overview of AlphaFold Database, AlphaFold Server, Protein Data Bank and UniProt.
- Retrieval of predicted and experimentally determined protein structures.
- Interpretation of pLDDT, PAE, ipTM and domain-confidence scores.
- Comparison of AlphaFold-predicted structures with experimental structures.
- Identification of active sites, binding pockets and functional residues.
- Protein cleaning, hydrogen addition, protonation and structural preparation.
- Limitations of predicted structures, flexible loops, missing residues and uncertain regions.
🛠️ Hands-on Activities
- Retrieve a disease-associated protein structure from AlphaFold Database or Protein Data Bank.
- Analyse confidence metrics and identify reliable and uncertain structural regions.
- Visualise the protein structure and identify potential ligand-binding pockets.
- Clean and prepare the selected protein structure for molecular docking.
🧰 Tools Covered: AlphaFold Database, AlphaFold Server, UniProt, Protein Data Bank, PyMOL or UCSF Chimera.
📅 Day 2: Molecular Docking and Protein–Ligand Interaction Analysis
Focus: Performing molecular docking and interpreting binding poses beyond docking scores.
- Principles of molecular docking and structure-based virtual screening.
- Retrieval of candidate compounds using PubChem, SMILES and SDF files.
- Ligand cleaning, protonation and geometry optimisation.
- Preparation of protein and ligand docking files.
- Selection of docking grids, search spaces and binding-site coordinates.
- Configuration of exhaustiveness and pose-generation parameters.
- Interpretation of docking scores, binding energies, RMSD and pose clustering.
- Analysis of hydrogen bonds, hydrophobic interactions, salt bridges and π-interactions.
- Limitations of docking scores and the importance of biological interpretation.
🛠️ Hands-on Activities
- Retrieve and prepare five to ten candidate compounds.
- Define the protein-binding site and docking search space.
- Perform molecular docking against the prepared protein target.
- Visualise and compare the top-ranked binding poses.
- Develop an initial interaction-based compound ranking.
🧰 Tools Covered: PubChem, AutoDock Vina through Google Colab, SwissDock or DockThor, PyMOL, Protein-Ligand Interaction Profiler.
📅 Day 3: ADMET Profiling, Toxicity Prediction and Lead Prioritisation
Focus: Integrating docking, drug-likeness, pharmacokinetic and toxicity results for rational lead selection.
- Drug-likeness and lead-likeness assessment using established medicinal chemistry filters.
- Evaluation of molecular weight, LogP, TPSA, solubility and molecular flexibility.
- Prediction of gastrointestinal absorption and oral bioavailability.
- Assessment of blood–brain barrier penetration and compound distribution.
- Analysis of CYP450 metabolism and drug–drug interaction risks.
- Evaluation of P-glycoprotein interactions, transporters and clearance.
- Prediction of hepatotoxicity, mutagenicity, carcinogenicity and acute toxicity.
- Assessment of cardiotoxicity and hERG-channel inhibition.
- Development of a weighted docking–ADMET lead-prioritisation matrix.
🛠️ Hands-on Activities
- Evaluate drug-likeness, lead-likeness and medicinal chemistry alerts.
- Predict absorption, distribution, metabolism and excretion properties.
- Assess toxicity classes, organ toxicity, mutagenicity and cardiotoxicity.
- Compare predictions obtained from multiple ADMET platforms.
- Build an integrated ranking matrix and select the final lead and backup compounds.
🧰 Tools Covered: SwissADME, pkCSM, ADMETlab 3.0, ProTox-II, PubChem, Microsoft Excel or Google Sheets
