In today’s rapidly advancing field of biomedical technology, designing biomolecules that can adapt effectively to the human body has long posed a major challenge for researchers. Traditional methods, constrained by experimental environments and limited efficiency, often fail to meet the urgent demands of precision medicine and next-generation drug development. Now, the “Bio-Artificial Intelligence” system PROTEUS (Protein Evolution Using Selection), developed by the University of Sydney, is providing a groundbreaking solution to this bottleneck. By simulating natural evolutionary mechanisms, PROTEUS enables rapid design and optimization of molecules within mammalian cells, driving transformative breakthroughs in drug discovery, gene editing, and more. Its core technology has been published in Nature Communications and independently validated by external laboratories, promising to reshape the landscape of biomedical research.
A Technological Leap: From Bacteria to Mammalian Cells
Traditional directed evolution technologies—recognized with the 2018 Nobel Prize in Chemistry—have historically operated mainly in bacterial systems. The advent of PROTEUS marks the first time this technology has been successfully extended to mammalian cells. This breakthrough makes it possible to directly design proteins and biomolecules that are better adapted to human physiological environments, opening entirely new avenues for research and application.
The key innovation behind this leap lies in the ingenious design of chimeric virus-like particles. The research team combined the shells and genomes of two different viruses to create a stable system capable of resisting “cheating” mutations. This system allows mammalian cells to process millions of molecular sequences in parallel, mimicking nature’s process of evolution by weeding out non-functional mutations and selecting optimal solutions. As a result, PROTEUS achieves an evolution-like process similar to machine learning within mammalian cells—compressing what once took years or even decades into just weeks, with remarkable gains in efficiency.
Broad Applications: Empowering Healthcare with New Possibilities
The potential applications of the PROTEUS system are vast, especially in accelerating drug development and advancing precision medicine.
In gene editing, PROTEUS has already improved the efficiency of CRISPR tools and designed new molecules capable of precisely silencing disease-causing genes. For example, through evolutionary screening, the team developed proteins with six times greater drug response sensitivity than conventional methods, laying a strong foundation for broader gene-editing applications.
In early cancer intervention, nanobodies designed by PROTEUS can detect DNA damage—a key hallmark of cancer—and accurately target the tumor suppressor protein p53 within the cell nucleus. This technology has been validated in both hamster and human cells, providing a powerful, high-efficiency tool for early cancer diagnosis and improving the chances of early detection and successful treatment.
Additionally, the algorithmic framework of PROTEUS can be extended to optimize the stability and targeting of mRNA therapeutics, enhancing the delivery of therapeutic genetic material and driving the evolution of mRNA-based treatments. It can also boost the functionality of most proteins and molecules, offering universal solutions for industrial enzyme engineering, biomanufacturing, and more—making PROTEUS a versatile platform for molecular design.
Outstanding Advantages: Surpassing Traditional Technologies
Compared to other AI drug design platforms—such as RFdiffusion and Chroma—PROTEUS demonstrates superior performance in design capability, efficiency, and diversity.
When it comes to handling long sequences, PROTEUS shows a significantly higher success rate than RFdiffusion, particularly for designing long-chain proteins with more than 300 amino acids—an area where pre-trained models often struggle.
In experimental validation, proteins generated by PROTEUS have a Cα-RMSD (root mean square deviation) of less than 2 Ångströms between the designed backbone and the predicted structure, fully meeting functional design standards and ensuring that its results are both reliable and practical.
In computational performance, PROTEUS’s graph-based triangulation techniques and multi-track interaction networks deliver a sampling speed over three times faster than RFdiffusion. Moreover, it can handle proteins up to 512 residues in length, breaking traditional model limits and further increasing its real-world applicability.
Currently, PROTEUS has been released as an open-source tool for the global research community, aiming to accelerate the development of next-generation enzymes, molecular tools, and therapeutics. The University of Sydney is collaborating with the Centenary Institute to apply this technology to preclinical development of gene-editing and mRNA drugs. As PROTEUS continues to evolve and expand across disciplines, bio-artificial intelligence is poised to spark a new revolution in precision medicine and biotechnology.
