Antimicrobial resistance is a growing global crisis, and Streptococcus pneumoniae — the bacteria behind deadly infections like pneumonia and meningitis — is becoming harder to treat with standard antibiotics. Researchers are now turning to artificial intelligence to speed up the search for solutions, and a new study published in Advanced Science shows just how promising that approach can be.

Scientists trained three distinct AI models — decision tree ensembles, graph neural networks, and sequence-based transformers — on datasets of molecules known to be active or inactive against drug-resistant bacteria. Together, these models screened nearly 7,000 existing, approved drugs for potential repurposing against S. pneumoniae. The diversity of models proved key: each complemented the others, making the combined approach more effective than any single model alone.

By the Numbers:

• ~7,000 candidate drugs screened by AI models
• 11 compounds selected for experimental validation
• 9 of 11 successfully inhibited S. pneumoniae growth
• 1 of the top 2 compounds remained effective against drug-resistant strains

Why it matters: Drug repurposing — using already-approved medications for new purposes — is faster, cheaper, and lower-risk than developing new drugs from scratch. With antimicrobial resistance claiming lives and straining healthcare systems worldwide, AI-driven repurposing offers a scalable, affordable path to new treatments urgently needed on the WHO's priority list.