17 October 2024
In an era where artificial intelligence (AI) is transforming every industry, the field of biotechnology is experiencing an exciting revolution with the emergence of AI-designed proteins. This groundbreaking technology is opening doors to new possibilities in drug development, industrial enzymes, and laboratory reagents, raising hopes for a transformative impact on medicine. However, the key question remains: can these AI-designed proteins actually work?
Recent protein-design competitions are offering a glimpse into the future, where AI-generated proteins could become game-changers in healthcare. These contests, which have emerged over the past year, are not only fueling rapid innovation but also helping scientists separate the functional proteins from the fantastical ones. By driving researchers to design and test new proteins, these competitions are quickly becoming the hotbed of scientific advancements that could revolutionize the biotech industry.
AI-Driven Protein Design: A Surge of Innovation
In August, Alex Naka, a protein engineer from California, joined the ranks of those testing the limits of AI-driven protein design. Fueled by coffee and cloud-based AI processors, Naka participated in a protein-design competition hosted by Adaptyv Bio, a Swiss biotech start-up. His goal? To create computer-generated proteins capable of blocking a cell receptor known to mutate in some tumors.
The competition was part of a growing number of AI-driven protein-design contests aimed at tackling pressing challenges in healthcare. With over 700 protein designs submitted, participants were hoping to unlock the potential of new proteins that could serve as more effective drugs, industrial enzymes, or reagents for research.
These contests follow in the footsteps of past scientific competitions like the Critical Assessment of Structure Prediction (CASP), which played a significant role in advancing protein-structure prediction. Notably, CASP spurred the development of AlphaFold, the AI tool that famously predicted the structure of nearly every known protein, earning its creators a Nobel Prize in Chemistry. Now, AI is poised to do the same for protein design.
Competitions: Accelerating Discovery in Protein Engineering
In the past year, at least five major protein-design competitions have emerged. They are not only drawing in seasoned scientists but also newcomers, making the field more accessible than ever before. Contests such as the Protein Engineering Tournament, Bits to Binders, and Liberum Bio’s Winter Protein Design Games are pushing participants to develop novel proteins with real-world applications.
These competitions offer more than just prizes—they provide valuable validation by testing the protein designs in laboratories. This direct feedback loop allows researchers to quickly iterate and refine their creations, saving months or even years of experimental work.
Despite the excitement, the field of AI-driven protein design faces challenges. Many scientists are generating proteins faster than they can be tested, leading to confusion over which designs are truly effective. Additionally, determining the criteria for success in protein design is far more complex than in protein-structure prediction, where the goal is simply to match the computational model with an actual protein structure.
AI’s Expanding Role in Protein Design
The stakes are high in the race to harness AI’s full potential in protein design. The technology is already being used to develop proteins that could improve treatments for diseases such as cancer, create environmentally friendly enzymes for industry, and even aid in the development of vaccines.
In some cases, AI is enabling non-specialists to participate in cutting-edge protein research. A participant from Iran, who had no formal experience in protein engineering, used a gaming computer to create his protein designs. This democratization of protein engineering is reminiscent of the early days of tech giants like Apple and Microsoft, which started in garages but went on to change the world.
“The fact that someone can enter a protein-design contest from their own kitchen using a laptop and AI tools is amazing,” says Julian Englert, CEO of Adaptyv Bio. “It’s opening the door for a whole new generation of innovators.”
Challenges Ahead: Failure and Success in the Lab
While AI can generate thousands of protein designs in a short amount of time, the real challenge lies in determining whether these proteins actually function as intended. In Naka’s case, despite his enthusiasm and high hopes, none of his ten top designs were successful in lab testing. But Naka’s experience is not unusual. Protein engineering is a field where failure is common, and only a small fraction of designs make it through the rigorous process of validation.
Of the 147 designs tested in Adaptyv Bio’s competition, only five successfully bound to the target molecule, an epidermal growth factor receptor (EGFR) implicated in many cancers. Despite the low success rate, experts consider this a significant achievement. Protein engineering has historically been a slow and tedious process, with even lower success rates in similar past efforts.
“This is part of the process in protein engineering—you have to be ready to fail a lot before you succeed,” Naka says. The winners of the Adaptyv Bio competition were Martin Pacesa and Lennart Nickel, structural biologists at the Swiss Federal Institute of Technology (EPFL) in Lausanne. Their success came from their open-source approach, making their designs and methods available to the wider community of protein researchers.
The Future of AI-Driven Protein Design
Looking ahead, the future of AI-driven protein design holds enormous potential. The field is already attracting attention from venture capitalists, with millions of dollars being funneled into biotech startups. Adaptyv Bio, for example, has launched a second competition to build on the success of its first, hoping to push the boundaries of protein design even further.
Meanwhile, other competitions are gearing up for their next rounds. The Protein Engineering Tournament plans to run its 2025 edition, and the Bits to Binders contest is currently testing 18,000 AI-designed proteins to determine their efficacy for use in cancer therapy. These competitions are not only accelerating innovation but also building a collaborative global community of scientists, engineers, and AI enthusiasts.
While the road ahead will be challenging, the promise of AI in protein design is undeniable. As researchers continue to refine their methods and improve the success rates of their designs, the possibilities for AI-designed proteins could revolutionize medicine, industry, and science in ways we can only begin to imagine.
In Naka’s words: “I’ll definitely be participating in similar events in the future. AI has lowered the barrier to entry, allowing a lot of new people to participate in protein design. The future is bright, and we’re just getting started.”
Read Now:Alaska’s Waterways Turn Rust Orange: Climate Change and Thawing Permafrost Blamed
