Machine learning ‘customized’ prescriptions promise to reduce risk of antibiotic resistance

Xinhua News Agency, Beijing, February 28 (Reporter Feng Yujing) Antibiotics play an important role in the treatment of bacterial infections, but they are a “double-edged sword” that also promotes bacteria to strengthen drug resistance. Using machine learning techniques and genome sequencing, an international team has developed an antibiotic prescribing algorithm that halved the risk of drug resistance in infection treatment, an international team reported in the US journal Science today.

The picture shows a file photo of an antibiotic counter in a pharmacy. (Photo by Xinhua News Agency reporter Guo Chen)

The focus of treating infection is to correctly match the resistance of antibiotics and pathogens. However, even if the matching is correct, there may still be Antibiotic resistance. One reason is that bacteria may mutate randomly during evolution to develop drug resistance, but this random process is difficult to predict and avoid.

A research team led by the Technion-Israel Institute of Technology found that drug resistance in most infected patients is not due to random mutations in pathogenic bacteria, but rather from resistance to prescribed antibiotics in the patient’s microbiome. Caused by rapid reinfection with another strain of sex. The researchers translated this finding into a therapeutic idea: The antibiotics used for treatment should not only match the resistance profile of the current pathogenic bacteria, but also match other bacteria in the patient’s microbiome that have the potential to replace the current pathogenic bacteria. match.

Utilizing eight-year records of microbiome data from more than 200,000 patients, the research team built a machine-learning algorithmic model to predict an individual’s risk of developing resistance to specific antibiotics. The researchers also trained the algorithm using large amounts of data on past antibiotic prescriptions for urinary tract infections and wound infections, allowing it to formulate personalized antibiotic treatment prescriptions. The study showed that the antibiotic prescribing algorithm halved the risk of developing antibiotic resistance in treatment.

“We found that patients’ susceptibility to antibiotics from past infections can be used to predict their risk of developing drug-resistant infections when they are re-treated with antibiotics.” Dr. Hugh Strassey explained.

The hope is that the study will provide doctors with better tools to personalize antibiotic regimens to improve outcomes and minimize the spread of drug-resistant pathogens, the researchers said.