An Introduction to Structure-based Small Molecule Drug Design

These materials are designed to help you learn the basics of computational small molecule drug design (small molecules are compounds smaller than proteins, typically containing around 20–70 heavy atoms). Using structure-based techniques, you will explore the key tools, concepts, and workflows through a series of interactive Jupyter notebooks. By the end you’ll be ready to explore how using computational tools can result in the discovery of your very own (virtual) drug candidates to cure Zika!

⚠️ Disclaimer

This repository is not a comprehensive guide to computational small molecule drug design. It reflects the perspectives and tool preferences of the authors and is intended for educational purposes only. These materials were developed with support from the Doctoral Training Centre at the University of Oxford.

0. Environment setup

Create SMDD_env environment

conda env create -f SMDD_env.yaml
conda activate SMDD_env

Install REINVENT4

For generating compounds we will use REINVENT4:

git clone https://github.com/MolecularAI/REINVENT4.git
cd REINVENT4
conda activate SMDD_env
python install.py cpu # or rocm6.2.4, cpu, mac, etc. depending on what OS you have, see REINVENT4 repo for help
reinvent --help

1. Final workshop goal

Design a compound that binds Zika virus NS2B–NS3 protease better than a known inhibitor. Justify your choice using property analysis, generative model scores, and docking interactions.

2. Notebook overview

# open notebooks
conda activate SMDD_env
jupyter lab

1_intro_to_rdkit.ipynb

Ligand-based analysis with RDKit: load molecules, calculate properties (MW, logP, TPSA), and filter for drug-likeness.

2_run_reinvent.ipynb

Generative design with REINVENT4: define a scoring function in TOML, generate candidate molecules, and triage the results.

3_zika_example_workflow.ipynb

Structure-based design: dock molecules into the Zika protease structure (PDB ID: 7I9O) and analyze protein–ligand interactions.

4_cure_zika.ipynb

Cure Zika!: combine all analyses to select 10 compounds and justify your choice.

3. Discovery pipeline

1. Hit understanding: Analyze the starting scaffold and its properties.
2. Idea generation: Use LibINVENT to generate new analogues with a custom scoring function.
3. Triage: Filter by toxicity and other properties (~100 molecules).
4. Docking: Dock into the Zika NS2B–NS3 protease structure (7I9O).
5. Compound selection: Select and justify 10 of the best compounds.