Building a Scaffold-Split Random Forest QSAR Co-Scientist with ChEMBL, RDKit, SHAP & BRICS to Discover EGFR Inhibitors
Decision Brief
What changedA tutorial on building an autonomous AI co-scientist for discovering EGFR C797S inhibitors.
Why it mattersFor drug discovery developers, it demonstrates a complete autonomous AI collaboration workflow: from data cleaning to model training, feature interpretation, and molecular generation and ranking.
Who should careAll AI builders
Affected stackNo specific stack identified
Source confidenceMedium · Reliable media or first-hand reporting
The tutorial uses ChEMBL and UniProt to analyze the EGFR C797S target, mining and cleaning pIC50 data from IC50 records. Molecules are standardized with RDKit, Morgan fingerprints computed, and a scaffold-split random forest QSAR model trained. SHAP explains potency-driving factors, and BRICS fragments are recombined to generate and rank new candidate molecules. For researchers in computer-aided drug design or pharma AI teams, this provides an end-to-end, reproducible open-source pipeline automating target parsing, data processing, model training, and candidate generation and ranking.
Summary basis: official / RSS sourceCompiled from the source scope noted above; the original remains authoritative.
Sources
- MarkTechPost
Fast research-paper and ML tooling summaries, useful for infra and agent updates.
- MarkTechPost