Google DeepMind's AlphaScience system has identified a novel class of antibiotic compounds effective against drug-resistant bacteria by combining AlphaFold protein structure predictions with generative chemistry models.
The Discovery
The AI system screened 12 million potential molecular structures against bacterial protein targets, identifying 236 candidates that were then synthesized and tested in laboratory conditions.
- Two compounds showed potent activity against MRSA (methicillin-resistant Staphylococcus aureus)
- One compound effective against carbapenem-resistant Enterobacteriaceae (CRE), a critical priority pathogen
- Novel mechanism of action — the compounds target a bacterial protein not addressed by existing antibiotics
- Time from AI discovery to lab validation: 8 months (traditional drug discovery: 4-6 years)
Significance
The WHO estimates that antimicrobial resistance will cause 10 million deaths annually by 2050 without new antibiotics. AI-accelerated drug discovery could dramatically compress the timeline for bringing new treatments to market, potentially saving millions of lives.
