Major project
Pharma & Life Sciences
Drug activity prediction for a target protein
Predict which molecules will be active against a disease target, to shortlist compounds for testing.
The brief
A drug-discovery team has thousands of candidate molecules and can only test a few hundred. Using public bioactivity data for one target protein, build a model that predicts activity from molecular structure, checks drug-likeness, and produces a ranked shortlist. Good looks like: good performance on new chemical scaffolds (not just similar molecules), a sensible shortlist, and a write-up a medicinal chemist would respect.
What you will build
- Pick a target and pull its bioactivity data
- Clean and standardise the molecules
- Compute fingerprints and descriptors
- Train models and evaluate with a scaffold split
- Filter for drug-likeness and rank a shortlist
- Report the findings like a short paper
Free, open data to build it with
- ChEMBL bioactivity database ↗ EMBL-EBI
- RCSB Protein Data Bank ↗ RCSB PDB
Check each dataset's licence before you publish your work. Your mentor gives your team its own version of the brief (data, scope or region), so every team's project is its own.