Ivan Kairatov is a leading figure in the biopharmaceutical industry, renowned for his ability to translate complex molecular research into actionable technological innovations. With an extensive background in research and development, Kairatov has closely followed the evolution of protein engineering, particularly the shift toward computational design. His insights are especially valuable today, as we witness a historic convergence of artificial intelligence and structural biology. In this discussion, we explore a groundbreaking achievement led by Professor Sangmin Lee and Nobel laureate David Baker, who have utilized AI to crack the code of viral architecture, creating artificial nanocages that could redefine the future of medicine.
The conversation explores the sophisticated design principles of protein nanocages, focusing on the concept of quasisymmetry as a means to overcome the limitations of traditional, perfectly symmetric models. We examine how the AI tool RFdiffusion allows scientists to manipulate protein curvature and angles, much like interlocking building blocks, to create massive dome-shaped shells. The discussion also highlights the practical verification of these designs through cryo-electron microscopy and the profound implications for targeted drug delivery, vaccine development, and the next generation of genetic therapies.
Traditional protein design often relies on perfect symmetry, yet natural viruses use quasisymmetry to build massive, complex shells. How did the research team bridge this gap to create artificial structures that mirror nature’s flexibility?
For decades, we were trapped in a design paradigm that demanded “perfect symmetry,” which is essentially like trying to build a cathedral using only square bricks of a single size. Nature is far more clever than that, and this research team finally leaned into the “messiness” of quasisymmetry to break those old constraints. By moving away from rigid, computationally derived patterns, they allowed a single protein component to act like a modular tile that can adjust its orientation based on its neighbors. They discovered that the secret lies in the delicate balance of curvature; if the proteins sit too flat, they form an endless sheet, but if they curve too sharply, they snap into a tiny, useless sphere. By precisely engineering a trimeric unit—a cluster of three proteins—they induced these building blocks to simultaneously occupy both pentagonal and hexagonal environments, allowing the structure to breathe and grow into a massive, stable shell.
The use of RFdiffusion seems to be a turning point in how we “program” proteins. Could you explain how this AI tool helped the team design connecting structures that allow a single protein to serve multiple roles within a single assembly?
RFdiffusion is truly the “engine under the hood” that made this architectural feat possible, acting as a generative architect that understands the physics of protein folding better than any human ever could. The team used this AI to design entirely novel connecting structures that act like flexible hinges between the protein units. Instead of guessing how to link these molecules, the AI simulated thousands of variations to find the exact geometry where a single protein could “fit” into different positions within the same shell. It is a bit like designing a 3D puzzle piece that changes its angle slightly depending on whether it is at the top of a dome or along the side. This level of precision ensured that the proteins didn’t just clump together randomly, but rather self-assembled into the elaborate, virus-like structures that we previously thought were impossible to create from scratch.
When the team looked through the cryo-electron microscopy, they saw structures ranging from 70 to 220 nanometers. What does this variation in size tell us about the future of custom-built bio-containers?
Seeing those images for the first time must have been an incredible “eureka” moment for the team, as the visual evidence confirmed that their mathematical models were alive and functioning in the real world. The fact that these proteins, produced in simple E. coli, could spontaneously form shells as small as a 70 nm “nano-soccer ball” and as large as a 220 nm giant is a testament to the versatility of the design. This size range is critical because it covers the exact dimensions needed to carry everything from small-molecule drugs to massive genetic payloads or complex enzymes. It suggests that we are moving toward a “bespoke” era of medicine where we can tailor the size of the delivery vehicle to the specific cargo it needs to protect. The sheer scale of the 220 nm structures, being more than three times the size of the smallest units, proves that we are no longer limited by the inherent size of the protein building block itself.
We often hear about “protein nanocages” as a promising platform for drug delivery. From your perspective, how does using a single, AI-designed protein simplify the path toward clinical applications like targeted gene therapy?
The beauty of a “one-component” system is its radical simplicity, which is a dream come true for large-scale manufacturing and regulatory approval. In the past, creating complex nanocages often required multiple different proteins to work in perfect harmony, which is a nightmare to stabilize and reproduce in a factory setting. By using a single, entirely AI-designed protein, the team has created a “plug-and-play” shell where the interior can be packed with genetic material and the exterior can be decorated with specific antigens. This creates a highly stable environment that protects fragile medicines from the body’s immune system until they reach their target. It’s a transformative leap because it turns the delivery vehicle from a passive box into an active, programmable tool that can be mass-produced with the consistency required for human clinical trials.
What is your forecast for the future of AI-driven protein architecture in the next decade?
In the next ten years, I expect we will stop talking about “artificial” proteins as a novelty and start seeing them as the primary scaffolding for all advanced therapeutics. We will likely see the development of even more sophisticated “template” systems, where internal scaffold proteins or even nucleic acids are used to dictate the exact, uniform size of every nanocage produced. This research, which earned Professor Sangmin Lee the rare honor of being a corresponding author and a co-author on two simultaneous Nature papers, is just the opening chapter of a much larger story. We are heading toward a future where we can design biological “machines” that are more efficient, more stable, and more targeted than anything found in the natural world. By 2035, the ability to program protein self-assembly with the precision of a computer circuit will likely be the standard method for treating everything from rare genetic disorders to global viral outbreaks.
