arstechnica.com
Stop the snake Researchers use AI to design proteins that block snake venom toxins It's a good example of how computer developments can be used for practical problems. John Timmer Jan 15, 2025 11:00 am | 26 Elapidae. Credit: Paul Starosta Elapidae. Credit: Paul Starosta Story textSizeSmallStandardLargeWidth *StandardWideLinksStandardOrange* Subscribers only Learn moreIt has been a few years since AI began successfully tackling the challenge of predicting the three-dimensional structure of proteins, complex molecules that are essential for all life. Next-generation tools are now available, and the Nobel Prizes have been handed out. But people not involved in biology can be forgiven for asking whether any of it can actually make a difference.A nice example of how the tools can be put to use is being released in Nature on Wednesday. A team that includes the University of Washington's David Baker, who picked up his Nobel in Stockholm last month, used software tools to design completely new proteins that are able to inhibit some of the toxins in snake venom. While not entirely successful, the work shows how the new software tools can let researchers tackle challenges that would otherwise be difficult or impossible.Blocking venomSnake venom includes a complicated mix of toxins, most of them proteins, that engage in a multi-front assault on anything unfortunate enough to get bitten. Right now, the primary treatment is to use a mix of antibodies that bind to these toxins, produced by injecting sub-lethal amounts of venom proteins into animals. But antivenon treatments tend to require refrigeration, and even then, they have a short shelf life. Ensuring a steady supply also means regularly injecting new animals and purifying more antibodies from them.Having smaller, more stable proteins that perform the same function would let us produce them in bacteria and could allow the generation of an antivenon that doesn't require refrigerationa careful consideration given that many snake bites occur in rural areas or the wilderness.The new work isn't meant to be a complete solution to the problem. Instead, it tackles a single type of toxic venom protein: the three-finger toxins, named after the physical structure that the proteins fold into. They're a major component of the venom of such infamous snakes as mambas, taipans, and cobras. Despite their relatively compact size, different members of the three-finger toxin family manage to produce two distinct types of damage. One group causes a general toxicity to cells, enabled by disruption of the cell membrane, while a different subset has the ability to block the receptor for a neurotransmitter.Since these two toxicities work through entirely different mechanisms, the researchers tackled them separately.Blocking a neurotoxinThe neurotoxic three-fingered proteins are a subgroup of the larger protein family that specializes in binding to and blocking the receptors for acetylcholine, a major neurotransmitter. Their three-dimensional structure, which is key to their ability to bind these receptors, is based on three strings of amino acids within the protein that nestle against each other (for those that have taken a sufficiently advanced biology class, these are anti-parallel beta sheets). So to interfere with these toxins, the researchers targeted these strings.They relied on an AI package called RFdiffusion (the RF denotes its relation to the Rosetta Fold protein-folding software). RFdiffusion can be directed to design protein structures that are complements to specific chemicals; in this case, it identified new strands that could line up along the edge of the ones in the three-fingered toxins. Once those were identified, a separate AI package, called ProteinMPNN, was used to identify the amino acid sequence of a full-length protein that would form the newly identified strands.But we're not done with the AI tools yet. The combination of three-fingered toxins and a set of the newly designed proteins were then fed into DeepMind's AlfaFold2 and the Rosetta protein structure software, and the strength of the interactions between them were estimated.It's only at this point that the researchers started making actual proteins, focusing on the candidates that the software suggested would interact the best with the three-fingered toxins. Forty-four of the computer-designed proteins were tested for their ability to interact with the three-fingered toxin, and the single protein that had the strongest interaction was used for further studies.At this point, it was back to the AI, where RFDiffusion was used to suggest variants of this protein that might bind more effectively. About 15 percent of its suggestions did, in fact, interact more strongly with the toxin. The researchers then made both the toxin and the strongest inhibitor in bacteria and obtained the structure of their interactions. This confirmed that the software's predictions were highly accurate.A mix of the three-fingered neurotoxin and the newly designed inhibitor was then injected into mice, where it provided complete protection (as long as there was five times more inhibitor than toxin). It even worked at a 10-fold excess when it was injected into the mouse 30 minutes after the toxin, which might better reflect real-world use of an antivenon.Mixed successAs mentioned above, a different group of three-fingered toxins can directly kill cells by disrupting their membranes. This class of toxin is made by spitting cobras, which means they can deliver the toxin to victims without the need to even bite them. Here, the researchers focused on the three fingers of the protein structure that gave this group its name. Repeating a similar process created inhibitors that interacted strongly with the three-fingered toxin and could potentially inhibit its activity.Unfortunately, when tested on actual mice, the inhibitors did not decrease the size of the skin lesions caused by the three-fingered toxin. This may indicate that we don't fully understand how these proteins disrupt membranes and could have potentially targeted the wrong region on them for inhibition. So the researchers stopped testing this inhibitor, though they could continue to work to identify others that target different areas of the protein.Even if they're successful, this work is mostly a proof of concept. Snake venoms typically contain a wide variety of toxins, and these experiments only targeted two of them. In addition, the proteins it produced worked well because they are highly specific. But that specificity means that an inhibitor designed against proteins in cobra venom might not work against the venom in a more distantly related snake.Still, the work shows that AI tools really can dramatically expand our options when it comes to intervening in biology. Without them, this work likely would have been stuck at the very first step, given that it was near-impossible to reason our way into identifying a protein structure that might interact with something like this toxin. And refining any initial ideas might have taken months to years of grunt work. It's hard to overstate just how radical a change the ability to do all this in software represents.Nature, 2025. DOI: 10.1038/s41586-024-08393-x (About DOIs).John TimmerSenior Science EditorJohn TimmerSenior Science Editor John is Ars Technica's science editor. He has a Bachelor of Arts in Biochemistry from Columbia University, and a Ph.D. in Molecular and Cell Biology from the University of California, Berkeley. When physically separated from his keyboard, he tends to seek out a bicycle, or a scenic location for communing with his hiking boots. 26 Comments
