Edit | Green Luo
Antibiotics have increased the average life expectancy by at least a decade. But antibiotics are not as effective as they once were, in part because of their widespread use.
"Health agencies around the world agree that we are entering a post-antibiotic era," explains Jacob Scott, MD, at the Cleveland Clinic. "If we don't change the way we track bacteria, by 2050 more people will die from antibiotic-resistant infections than from cancer."
Researchers at the Cleveland Clinic have developed an artificial intelligence (AI) model that can determine the optimal combination and timeline of drugs to treat bacterial infections based solely on how quickly bacteria grow under specific perturbations.
相关研究以「Reinforcement learning informs optimal treatment strategies to limit antibiotic resistance」为题，于 2024 年 2 月 23 日，发布在《Proceedings of the National Academy of Sciences》（PNAS）上。
Paper link: https://doi.org/10.1073/pnas.2303165121
Bacteria replicate quickly, producing mutated offspring. Overuse of antibiotics gives bacteria a chance to practice producing mutations that resist treatment. Over time, antibiotics kill all susceptible bacteria, leaving only stronger mutants that antibiotics can't.
One strategy doctors use is antibiotic rotation (cycling) to modernize the way we treat bacterial infections. Healthcare providers rotate different antibiotics for specific periods. Switching to a different drug can shorten the time it takes for bacteria to become resistant to any class of antibiotics. Rotation can even make bacteria more susceptible to other antibiotics.
"Drug rotation shows great promise in effectively treating disease," said Dr. Davis Weaver, a medical student and first author of the study. "The problem is that we don't know the best way to do it. There is no uniform standard between hospitals for which antibiotics to use, for how long, and in what order."
Study co-author Jeff Maltas, Ph.D., a postdoctoral researcher at the Cleveland Clinic, uses computer models to predict how a bacterium's resistance to one antibiotic will make it less resistant to another. He collaborated with Dr. Weaver to investigate whether data-driven models could predict drug rotation regimens that would minimize antibiotic resistance and maximize antibiotic susceptibility, despite the stochastic nature of bacterial evolution.
Figure: Schematic evolutionary simulations and tested optimization methods. (Source: Paper)
Dr. Weaver spearheaded the application of reinforcement learning to a drug rotation model that teaches computers to learn from mistakes and successes to determine the best strategy to complete a task. This study is one of the first to apply reinforcement learning to antibiotic rotation therapy. Weaver and Maltas said.
"Reinforcement learning is an ideal approach because you just need to know how fast the bacteria are growing, which is relatively easy to determine," Dr. Weaver explains. "There is also room for human change and error. You don't need to measure growth rates perfectly, down to the millisecond, every time."
Figure: Performance of the RL agent in a simulated E. coli system. (Source: Paper)
The research team's AI was able to identify the most effective antibiotic rotation programs to treat multiple strains of E. coli and prevent drug resistance. Dr. Maltas says research shows that AI can support complex decisions, such as calculating antibiotic treatment regimens.
In addition to managing infections in individual patients, the team's AI model can also inform hospitals on how to treat infections holistically, Dr. Weaver explains. He and his research team are also working to expand their work beyond bacterial infections to other deadly diseases.
"The idea is not limited to bacteria, it can be applied to anything that can create treatment resistance," he said. "In the future, we believe that these types of AI can also be used to treat drug-resistant cancers."
References: https://medicalxpress.com/news/2024-04-ai-treatments-superbugs.html