Molecule property and toxicity predictor
Inspired by publicly reported work at Novartis
Predict drug-like properties and toxicity of molecules from their structure, to help choose which to test.
The brief
Novartis has described using data and AI to speed up drug research. Build one tool a discovery team uses: predict properties such as solubility and toxicity from a molecule's structure, compare fingerprint models with graph neural networks, and rank candidate molecules for testing with uncertainty. Good looks like: scaffold splits rather than random ones, results on standard benchmarks, uncertainty that is honest, and a small screening app that ranks molecules.
Read the case first
What Novartis has said in public about this kind of work, on its own pages:
Make it your own
Focus on a disease area that needs cheap medicines: tuberculosis or malaria compounds from ChEMBL. Add one rule your screening must respect (cost to make, or a known toxic group) and show how the ranking changes.
AITeky has no link with Novartis, which has not reviewed or endorsed this brief. The brief only draws on what the company has published, uses open data rather than its data, and is a learning exercise: don't present your project as Novartis's work or system.
What you will build
- Load benchmark datasets and molecules (SMILES)
- Compute fingerprints and descriptors with RDKit
- Train gradient boosting and a graph neural network
- Evaluate on scaffold splits
- Add uncertainty and rank candidates
- Build a screening app and write a model card
Free, open data to build it with
Sign up free to get the data and links
It takes a minute with your email. AiTeky Pass (₹999/month or ₹9,999/year) is free for your first 30 days.