How AI could help scientists find viruses to fight drug-resistant infections
Microsoft’s Discovery app assisted researchers in building a cloud-based bioinformatics pipeline to identify bacteriophages against drug-resistant bacteria, accelerating phage therapy workflows without claiming treatment results.
Drug-resistant infections pose a growing global health threat, and bacteriophages—viruses that infect bacteria—offer a potential solution when antibiotics fail. However, locating effective phages in vast environmental datasets is challenging due to genetic variability and bacterial defenses. Researchers used the Microsoft Discovery app to create a structured bioinformatics workflow that ranks phage candidates by their predicted ability to infect and kill specific drug-resistant bacteria, integrating multiple analytical methods and scientific evidence.
The team combined expertise in genomics, bioinformatics, and cloud computing with agents in the Discovery app to design a reproducible pipeline. The app translated scientific goals into a task tree, linking research steps to curated data sources, analytical tools, and validation criteria. Scientists maintained control over the workflow, allowing iterative refinement based on expert review and new evidence, while the app managed task sequencing and traceability of research outputs.
The workflow begins by analyzing a target bacterium’s genetic profile to identify antibiotic-resistance markers, potential phage entry points, and bacterial defenses. It then searches environmental data for phages that could overcome these barriers, combining signals such as DNA similarity, known phage–bacterium relationships, and predicted bacterial surface interactions to rank candidates for further evaluation.
The pipeline excludes temperate phages to avoid risks like horizontal gene transfer and includes safety checks for engineered phage options. While computational predictions narrow the field, laboratory experiments remain essential to confirm effectiveness, measure resistance development, and assess safety. The result is a structured, reviewable process that focuses resources on the most promising candidates while maintaining rigorous scientific oversight.