AI designs proteins that kill antibiotic-resistant E. coli

Australian scientists have used artificial intelligence to create custom proteins that can kill antibiotic E. coli bacteria by starving basic nutrients.

Breakthroughs show how AI can quickly design biological weapons to evolve superbs that are resistant to traditional antibiotics, potentially changing how we combat dangerous infections.

The study, published in Nature Communications, represents the first time that Australian scientists have used AI to generate ready-made therapeutic proteins. Often, it is done in a few weeks using computational design tools that usually take decades to develop.

Targeted bacterial iron theft

The team is committed to disrupting the key survival mechanisms used in the pathogen E. coli. These bacteria steal iron from human hemoglobin using a specialist transporter called Chua, which specifies heme (the iron-containing component in the blood) from the host protein.

Iron is crucial for bacterial growth and survival, but the human immune system isolates iron without iron as a defense mechanism. To overcome this, E. coli evolved Chua into a molecular pirate that quickly extracted heme from hemoglobin without forming a stable complex, making it difficult for traditional methods to stop.

Using structural modeling and AI-driven protein design tools, the researchers created artificial proteins that bind to Chua’s extracellular circuits, physically blocking their ability to interact with hemoglobin and steal iron.

Rapid design and testing process

The team generated about 20,000 potential protein designs using RFDiffusion and ProteinMPNN (available AI tools). After the calculation screening, they selected 96 candidates for laboratory testing.

The results exceeded expectations. Several proteins designed exhibit significant efficacy:

• Protein G7 achieves 42.5 namore inhibits bacterial growth
• Various designs work at sub-100 nanomolar concentration
• Proteins specifically block heme in hemoglobin without affecting other bacterial functions
• The frozen structure almost perfectly confirms the matching of design and computational predictions

This represents a much higher success rate than traditional drug discovery methods, where most compounds fail during development.

Accurate positioning strategy

The designed proteins are controlled by competitive inhibition – they bind to the same site on Chua that usually interacts with hemoglobin, thus preventing bacteria from entering this critical iron source. Importantly, proteins do not interfere with free iron transport, meaning they are specifically targeted at the mechanisms that steal hemoglobin.

Biolayer interferometry studies show that artificial proteins bind to Chua and dissociation constants of 71 to 127 nanometers, which coincide with the natural CHUA-Hemoglobin interaction. This high affinity ensures effective competition for the bound sites.

The study confirms the specificity of the mechanism by demonstrating that proteins have no effect when providing an alternative iron source.

Structural Verification

The frozen structure of the Chua protein complex showed significant consistency with the calculation predictions, with a root mean square deviation of only 0.6Å. This close match demonstrates the accuracy of current AI protein design tools and validates structural models for understanding bacterial iron collection.

These structures suggest that the designed proteins occupy the predicted hemoglobin binding sites while simultaneously leading to minimal interference with other CHUA functions, explaining their selective inhibition of heme extraction.

A broader meaning

According to Dr. Rhys Grinter, co-led by the study, this approach can revolutionize antibacterial development. “These new approaches in deep learning can effectively design proteins with specific characteristics and functions from scratch, reducing costs and accelerating the development of novel protein binders and engineered enzymes,” he explained.

The study established Australia’s first AI protein design platform that enables the production of therapeutic proteins from scratch. Doctoral student Daniel Fox, who has conducted a lot of experimental work, highlighted the potential of democratization: “It is important to democratize protein designs so that the world can leverage these tools.”

Now, the platform can design proteins for a wide range of applications other than antibiotics, including medicines, vaccines and diagnostic tools. This makes Australia tied with the United States and China with AI-AI-kinetic protein design capabilities.

As antibiotic resistance represents a growing global health crisis, this AI-driven approach provides hope for the development of new antibiotics that work through novel mechanisms – potentially keeping bacterial evolution ahead of the spectrum and providing sustainable solutions for the treatment of dangerous infections.

Related