Reviewed by Lexie CornerApr 25 2024
Researchers at Cleveland Clinic have created an artificial intelligence (AI) model capable of identifying the optimal combination and timing for prescribing drugs to combat bacterial infections, relying solely on bacterial growth rate under specific conditions. The findings were published in PNAS.
The research was led by Jacob Scott, MD, Ph.D., and his team in the Theory Division of Translational Hematology and Oncology.
Thanks to antibiotics, the average US lifespan has been reported to have increased by nearly ten years. While treatments have historically reduced fatality rates for what are now considered minor health issues, such as certain cuts and injuries, antibiotics are becoming less effective due to their widespread use.
Health agencies worldwide agree that we are entering a post-antibiotic era; if we do not change how we go after bacteria, more people will die from antibiotic-resistant infections than from cancer by 2050.
Jacob Scott, Translational Hematology Oncology Research, Cleveland Clinic
Bacteria multiply swiftly, giving rise to mutant progeny. Overuse of antibiotics allows bacteria to evolve resistance to treatment through practice. The stronger mutants that the antibiotics are unable to destroy are all that remain after the antibiotics gradually eradicate all the susceptible bacteria.
Antibiotic cycling is one tactic doctors are employing to update the way they treat bacterial infections. Throughout predetermined intervals of time, medical professionals switch between various antibiotics. When they are switched between different medications, bacteria have less time to develop resistance to any one class of antibiotic. Cycling may increase a bacteria's susceptibility to additional antibiotics.
Drug cycling shows a lot of promise in effectively treating diseases; the problem is that we do not know the best way to do it. Nothing's standardized between hospitals for which antibiotic to give, for how long, and in what order.
Davis Weaver, Ph.D., Study First Author and Medical Student, Cleveland Clinic
Co-author of the study, Jeff Maltas, Ph.D., a Postdoctoral Fellow at Cleveland Clinic, forecasts how antibiotic resistance in a bacterium will reduce its resistance to another using computer models.
With Dr. Weaver, he set out to determine whether data-driven models could be used to forecast drug cycling regimens that maximize antibiotic susceptibility and minimize antibiotic resistance despite the stochastic nature of bacterial evolution.
The effort to integrate reinforcement learning, which teaches a computer to learn from its failures and successes to determine the optimal course of action to accomplish a task, into the drug cycling model was spearheaded by Dr. Weaver. According to Dr. Weaver and Dr.Maltas, this study is among the first to apply reinforcement learning to antibiotic cycling regimens.
Reinforcement learning is an ideal approach because you just need to know how quickly the bacteria are growing, which is relatively easy to determine; there is also room for human variations and errors. You do not need to measure the growth rates perfectly down to the exact millisecond every time.
Davis Weaver, Ph.D., Study First Author and Medical Student, Cleveland Clinic
The AI employed by the research team determined the best antibiotic cycling strategies for treating various E. Coli strains while avoiding drug resistance. According to Dr. Maltas, the study demonstrates how AI can assist with complicated decision-making, such as scheduling antibiotic treatments.
Dr. Weaver explains that the AI model developed by the team can help hospitals treat infections more broadly in addition to managing the infection of a specific patient. He is also attempting to broaden the scope of his research beyond bacterial infections to include other fatal illnesses.
Dr. Weaver said, “This idea is not limited to bacteria; it can be applied to anything that can evolve treatment resistance; in the future, we believe these types of AI can be used to manage drug-resistant cancers, too.”
Journal Reference:
Weaver, D. T., et al. (2024). Reinforcement learning informs optimal treatment strategies to limit antibiotic resistance. Proceedings of the National Academy of Sciences of the United States of America. doi.org/10.1073/pnas.2303165121.