AI-Generated Protein Shows Promise Against Antibiotic-Resistant E. Coli

AI-aided design offers a new avenue in protein development

A team of Australian researchers has developed a synthetic protein capable of targeting antibiotic-resistant strains of Escherichia coli (E. coli). The work, published in Nature Communications, was conducted using a local AI Protein Design Platform and marks a first for an Australian-led initiative in this area.

The research was led by scientists from the Monash Biomedicine Discovery Institute and the Bio21 Institute at the University of Melbourne. Their platform uses advanced deep learning tools to design proteins from scratch, rather than modifying those found in nature. The newly created protein, termed "de novo-1", binds to a heme-related target in E. coli, preventing the bacteria from acquiring the iron it needs to survive.

Building on recent developments in AI-based protein design

The Australian platform draws on global developments in AI-assisted protein engineering. It follows on from tools created by researchers including Nobel Laureate David Baker, whose work has laid the foundation for AI-driven protein synthesis.

Using tools such as Bindcraft and Chai, researchers are now able to model how protein sequences fold into three-dimensional shapes and predict how those shapes interact with specific molecular targets. In this case, the researchers aimed to interrupt a mechanism called “heme piracy”, which pathogenic strains of E. coli use to extract iron from host cells.

The AI-generated protein was validated experimentally in cell-based models, demonstrating its potential to function as an antimicrobial agent. The study focused on how this protein binds and blocks a key heme receptor, effectively starving the bacteria of iron and halting its growth.

Accelerating drug discovery pipelines

Traditional protein drug development typically relies on modifying existing proteins derived from natural sources. This process can take years of iterative design and testing. In contrast, de novo protein design allows researchers to specify the structure and function they want, then use AI algorithms to generate candidates that meet those criteria.

The team notes that this approach reduces the time and cost associated with therapeutic protein development. It also opens the door to creating highly specific molecules that are less likely to produce side effects or trigger immune responses.

National capability with global relevance

While similar AI protein design platforms exist in the United States and China, this marks the first such system established in Australia. The program behind the study is designed to be collaborative and adaptable, integrating new AI tools as they become available.

The researchers emphasize that the AI-driven design process is accessible to other scientific teams via publicly available software, supporting global efforts to develop new proteins for a range of applications.

Although this study was conducted using cell-based systems and has not yet moved into clinical or animal models, it demonstrates a promising approach to addressing antibiotic resistance using next-generation biotechnologies.

Reference: Fox D, Grinter R, Knott G, et al. Inhibiting heme piracy by pathogenic Escherichia coli using de novo-1 designed proteins. Nat Commun. 2025. doi:10.1038/s41467-025-60612-9 

This content includes text that has been generated with the assistance of AI. Technology Networks' AI policy can be found here.