The research, published in Nature Biomedical Engineering, marks an important step forward in creating safer and more effective treatments.
The research team employed a large language model (LLM), similar to the technology behind ChatGPT, to redesign Protegrin-1. This powerful antibiotic, produced naturally by pigs, was effective in killing bacteria, but was previously too toxic for human use.
By modifying Protegrin-1, the researchers aimed to preserve its antibacterial properties while eliminating its harmful effects on human cells.
To achieve this, the team generated more than 7,000 variations of Protegrin-1 using a high-throughput method, allowing them to quickly identify which modifications could improve safety. They then used LLM to evaluate these variations for their ability to selectively target bacterial membranes, effectively kill bacteria, and avoid damaging human red blood cells. This AI-guided approach led to the creation of a refined version known as bacterially selective Protegrin-1.2 (bsPG-1.2).
In preliminary animal tests, mice treated with bsPG-1.2 and infected with multidrug-resistant bacteria showed a significant reduction in bacterial levels in their organs within six hours. These promising results suggest that bsPG-1.2 could potentially advance to human trials.
Claus Wilke, professor of integrative biology and co-senior author of the study, highlighted the transformative impact of AI on drug development.
“Large language models are revolutionizing protein and peptide engineering, allowing new drugs to be developed and existing ones improved more efficiently. “This technology not only identifies potential new treatments but also accelerates their path to clinical application,” Wilke said.
The breakthrough underscores how AI is being leveraged to address critical health challenges.
The research team employed a large language model (LLM), similar to the technology behind ChatGPT, to redesign Protegrin-1. This powerful antibiotic, produced naturally by pigs, was effective in killing bacteria, but was previously too toxic for human use.
By modifying Protegrin-1, the researchers aimed to preserve its antibacterial properties while eliminating its harmful effects on human cells.
To achieve this, the team generated more than 7,000 variations of Protegrin-1 using a high-throughput method, allowing them to quickly identify which modifications could improve safety. They then used LLM to evaluate these variations for their ability to selectively target bacterial membranes, effectively kill bacteria, and avoid damaging human red blood cells. This AI-guided approach led to the creation of a refined version known as bacterially selective Protegrin-1.2 (bsPG-1.2).
In preliminary animal tests, mice treated with bsPG-1.2 and infected with multidrug-resistant bacteria showed a significant reduction in bacterial levels in their organs within six hours. These promising results suggest that bsPG-1.2 could potentially advance to human trials.
Claus Wilke, professor of integrative biology and co-senior author of the study, highlighted the transformative impact of AI on drug development.
“Large language models are revolutionizing protein and peptide engineering, allowing new drugs to be developed and existing ones improved more efficiently. “This technology not only identifies potential new treatments but also accelerates their path to clinical application,” Wilke said.
The breakthrough underscores how AI is being leveraged to address critical health challenges.
