2025: The Year of the Snake – How AI is Shaping Cheaper, Faster, and More Effective Antivenoms
As the Lunar New Year ushers in the Year of the Snake, GeneOnline is exploring exciting advancements in snakebite treatment, fueled by AI-powered protein design. This breakthrough technology offers a promising new approach to creating faster, more cost-effective, and more accessible antivenoms. Unlike traditional methods that rely on animals for antibody production, this technology uses computer-driven models to create small, synthetic proteins that target venom toxins. These proteins could be cheaper to manufacture, easier to store, and more efficient in reaching targeted areas within the body. The method also holds the promise of addressing venom toxins that currently have no available antivenom.
AI Designs Synthetic Antivenoms, Promising Cheaper, Faster, and More Effective Snakebite Treatments
Each year, an estimated 5.4 million people are bitten by snakes worldwide, with 1.8 to 2.7 million cases resulting in envenomings, according to the WHO. Venomous snakebites lead to approximately 81,410 to 137,880 deaths annually, while three times as many individuals suffer amputations and permanent disabilities. These bites can cause severe complications, including paralysis that impedes breathing, bleeding disorders that result in fatal hemorrhages, kidney failure, and tissue damage leading to irreversible disability and limb loss. Agricultural workers and children are particularly vulnerable, with children often experiencing more severe effects due to their smaller body mass.
Traditionally, antivenoms—crucial treatments for snakebites—are produced by injecting animals with venom to generate antibodies, a process that hasn’t changed in more than 100 years. However, a new study published in Nature suggests that artificial intelligence (AI) could offer a faster, cheaper, and more effective alternative for creating antivenoms.
The breakthrough comes from using synthetic antivenoms designed by AI, specifically a program called RFdiffusion. Developed by Nobel laureate David Baker and colleagues at the University of Washington, RFdiffusion designs small proteins known as “binders” that can specifically target venom toxins. This AI-driven process marks a dramatic shift from the traditional method, where animals’ immune systems do the work of producing antibodies. According to Baker, “It’s really this whole change from letting the animals’ immune system do the work to now being able to do it very intentionally on the computer.”
AI-Designed Proteins Show Promise in Neutralizing Snake Venom Toxins, Outperforming Traditional Antibodies
Inspired by a 2022 preprint from David Baker’s lab detailing AI-designed proteins that adhered to specific molecules, Timothy Jenkins, a medical biotechnologist at the Technical University of Denmark, saw an opportunity to test this technology with snake venoms. Although initially skeptical about the results, Jenkins decided to explore the potential by reaching out to Baker to see if his team could tackle the challenge.
Interestingly, Baker’s postdoc, Susana Vázquez Torres, had already been considering applying these AI-designed proteins to neutralize snake venom toxins. Although Baker had initially thought the task was too challenging, he was now ready to give it a go. Vázquez Torres eagerly began testing the proteins, quickly dispelling Jenkins’ doubts. “It was pretty fun to see the change in mindset,” Jenkins remarked.
The first batch of AI-designed proteins (called binders) successfully neutralized dangerous toxins from snakes, like the three-finger toxins found in cobras. These binders were even more effective than traditional antibodies in protecting cells and mice from the toxins. They worked whether mixed with venom beforehand or injected shortly after the venom was delivered, with no apparent side effects.
AI-Designed Proteins Show Promise in Neutralizing Cobra Venom, But Testing in Humans Remains the Next Step
Baker’s team initially trained the AI model on all known protein structures and their amino acid sequences. These amino acids are the molecular building blocks that fold into a protein’s 3-D shape. By computationally breaking down these shapes, the model learned how to reconstruct a protein from its individual components. It is similar to understanding how to build a car engine by first dismantling one.
Baker and Jenkins tasked the AI with designing proteins to target venom toxins. They synthesized the proteins in the lab. These custom proteins acted like magnetic caps that blocked the toxins from attaching to cells. It is similar to how a cap can prevent a key from fitting into a lock.
To test their effectiveness, the team injected 20 mice with a lethal dose of cobra toxins. They injected either 15 minutes after or simultaneously with the synthetic proteins. Remarkably, all the mice survived. “We were very, very excited about this,” Jenkins shares, marking a major success for the proteins.
The next challenge is to refine these proteins into a product that can be tested on humans. They must ensure the proteins are safe and don’t bind unexpectedly in human tissues.
Jenkins acknowledges that the study is just the beginning of exploring the potential of this new technology. “It was very much just proving that this extremely new technology works,” he adds.
Advancing Snakebite Treatment: AI-Designed Proteins vs. Traditional Antivenoms
Venomous snakes inject a range of toxins through their bites, with three-finger toxins among the most dangerous. These toxins paralyze muscles, halting both the heart and breathing. While antivenoms exist, they rely on outdated methods. Jenkins says these methods lack innovation due to limited financial interest. “There’s not a lot of money in it, so not a lot of innovation has been attracted,” Jenkins explains.
Current antivenom production involves extracting venom by milking snakes, a dangerous process Jenkins compares to handling “a live hand grenade.” The venom is injected into large animals like horses, and their antibodies are harvested. These antibodies neutralize the venom toxins when administered to snakebite victims. However, traditional antivenoms are expensive and time-consuming. Scientists are now exploring new alternatives. One promising approach is scanning a library of lab-made antibodies to find those that target specific toxins.
With the power of AI, scientists can design proteins from scratch to target these toxins quickly and cost-effectively. Jenkins and Baker collaborated to develop custom proteins using RFdiffusion, a generative AI model. Unlike image-generating AIs, RFdiffusion designs proteins that precisely match the toxins scientists aim to target. This offers a potential breakthrough in snakebite treatment.