Researchers in India and France built an AI system to guide antibiotic combos against resistant infections.
Researchers at IIIT-Delhi joined engineers at Inria Saclay in France to study drug resistance. The team traced how bacteria in hospitals no longer respond to single antibiotics. They set out to use data on past treatments, bacterial genomes, and drug structures to offer doctors new options.
The project began under a wider India-France partnership between Deep Light in Delhi and CentraleSupélec. Professor Angshul Majumdar of IIIT-Delhi and Dr Emilie Chouzenoux led the effort. They invited engineer Stuti Jain and graduate students Kriti Kumar and Sayantika Chatterjee to build the system.
“This is an excellent example of how AI and international collaboration can come together to solve real-world medical challenges, and our method makes it possible to use existing knowledge more effectively and opens the door to smarter, faster responses to antimicrobial resistance (AMR),” Prof. Majumdar told PTI.
The team gathered clinical records from major hospitals in India. They matched those decisions with the genome data of bacteria that resist drugs. They fed the chemical profiles of antibiotics into a machine-learning model. The AI then ranked possible drug combinations.
In tests, the system tackled multi-drug-resistant strains such as Klebsiella pneumoniae, Neisseria gonorrhoeae, and Mycobacterium tuberculosis. It showed how pairing two or more existing antibiotics could overcome resistance where single drugs failed.
Prof. Majumdar pointed to a case at AIIMS Kalyani. A young patient’s hip implant became infected. Doctors tried every last-resort antibiotic without success. He described how the infection “was a simple muscle infection, but it became untreatable. The patient from a poor family was left bedridden and without options.” The AI would have proposed several therapy cocktails based on the strain’s genome, offering a menu of paths for the medical team.
AMR arises when bacteria adapt to drugs meant to kill them. Overuse of antibiotics fuels that process. “We often take antibiotics even for viral infections, which don’t require them. Over time, bacteria evolve and adapt. As a result, even simple infections like urinary tract infections or wounds can become resistant to treatment,” Prof. Majumdar added.
The system can list five or ten treatment options for a given strain. Doctors can compare those choices and select one that fits their patient’s needs and local drug availability. The research team plans to refine the model further and train it on fungal pathogens and drug resistance in chronic diseases like hypertension.
The researchers aim to embed the AI tool into hospital records and public health networks. They see its first home in regions where lab diagnostics are scarce and AMR is widespread. A data-driven guide could help clinicians act fast when every hour counts.
The work underlines a broader goal: to bring AMR solutions to Southeast Asia, Africa and Latin America, where funding gaps slow new drug development. By using existing antibiotics more smartly, the AI may buy time for new treatments to reach the market.
As drug-resistant infections rise, the team hopes their system will serve as a second opinion for clinicians. A tool that learns from real-world cases and bacterial evolution could change how the world fights superbugs.
