University of Zurich researches resistance to antibiotics with AI

Antibiotic resistance is increasing worldwide and is a serious problem in the treatment of bacterial diseases. The precise identification of resistance is necessary in order to be able to develop antibiotics that react effectively. According to a press release, researchers at the University of Zurich(UZH) have now used a powerful OpenAI model supported by artificial intelligence to test the effectiveness or ineffectiveness of antibiotics.

The GPT-4 model is used in agar diffusion tests, such as the Kirby-Bauer disk diffusion test. Microorganisms are placed on a nutrient agar and their growth behavior is observed when antibiotics are added. Using the program developed with the help of GPT-4, the team led by Adrian Egli, Head of Research at the Institute of Medical Microbiology at UZH, tested hundreds of bacteria. Many resistances to vital antibiotics have been identified. The model used complied with the strict guidelines of the European Committee on Antimicrobial Susceptibility Testing(EUCAST)

As the researchers admit, the new system achieved good results for certain types of resistance, but for a number of others it marked samples as resistant even though they were not. Nevertheless, the AI model can supplement and support the analysis process of medical professionals due to its high diagnostic speed, as a study shows. This calls for further standardization of the tests.

"Our study is an important first step, but we are still a long way from replacing human expertise," Adrian Egli is quoted as saying in the press release. "Rather, we see AI as a complementary tool that can support microbiologists in their work." ce/ww