Researchers have used artificial intelligence to screen nearly 7,000 candidate drugs for activity against Streptococcus pneumoniae, a bacterial pathogen responsible for life-threatening infections including pneumonia and meningitis, and have identified one compound effective even against drug-resistant strains. The study is published in Advanced Science.
Antimicrobial resistance in S. pneumoniae is a recognised global health priority. The World Health Organization lists it among pathogens for which new antibiotic development is urgently needed. Drug repurposing, which involves identifying new applications for compounds with established safety profiles, offers a faster and more cost-effective pathway than de novo drug development, and computational screening has made this process considerably more efficient.
The research team trained three distinct machine learning algorithms on a dataset of molecules with demonstrated activity against S. pneumoniae and a larger dataset of molecules inactive against a separate drug-resistant bacterium. The three model types, comprising ensembles of decision trees, graph neural networks, and sequence-based transformers pretrained on hundreds of millions of molecules, were applied in combination to screen the full drug library. The authors found that using multiple complementary AI architectures improved candidate selection compared with any single model used in isolation.
From the initial screening, 11 compounds were selected for experimental laboratory validation. Nine demonstrated the ability to inhibit S. pneumoniae growth. Of the two most potent candidates, one retained activity against drug-resistant strains of the pathogen, representing a potentially significant finding in the context of antimicrobial resistance.
The authors propose that AI-guided phenotypic drug repurposing is a viable and scalable strategy for identifying effective compounds against priority pathogens, particularly where existing active molecules can serve as training data.
Source: Wiley. Artificial intelligence-guided phenotypic drug repurposing against Streptococcus pneumoniae. Advanced Science (2026). DOI: 10.1002/advs.76959
