Scientists develop AI-designed snake antivenom

A snakebite can be a death sentence in many parts of the world. Every year, millions of people experience venomous bites, yet treatment remains outdated, expensive and difficult to access – especially in remote regions. Traditional antivenoms, derived from animal plasma, are costly to produce and often fail to neutralise key toxins effectively. Frustrated by these limitations, scientists have turned to AI to design a more rapid, less expensive and more precise alternative: synthetic proteins engineered to counteract venom at the molecular level.

“This is more than just an improvement – it is a whole new way to treat snakebites. Instead of using immune proteins taken from animals, scientists are now designing custom-made molecules that work more rapidly and better and cost less. These synthetic proteins are stable, easy to produce and highly effective. The next step is to scale up production, create treatments tailored to different regions and eventually develop a simple injection that could save thousands of lives each year,” explains a lead researcher, Timothy Jenkins, Associate Professor at the Department of Biotechnology and Biomedicine, Technical University of Denmark, Kongens Lyngby.

Molecules often fail

Snakebite envenoming is a serious but often overlooked public health problem, especially in low-resource regions such as sub-Saharan Africa, south Asia, Papua New Guinea and Latin America. Each year, more than 2 million people are bitten, with 100,000 dying and 300,000 becoming permanently disabled. Nevertheless, treatment has barely changed for decades. Traditional antivenoms come from animal plasma, require complex medical support and are costly and difficult to distribute in remote areas.

“Snakebite treatment is incredibly complex – so many different toxins, so many different snakes. If you need antibodies for each species in a region, it is a massive undertaking in both time and resources. We have made huge strides using traditional antibody technologies, but it is still a slow and expensive process. I am always looking for ways to speed it up and make it more affordable so that the final product is accessible to those who need it most,” says Timothy Jenkins.

One challenge with conventional antivenoms is their limited ability to neutralise nerve-damaging toxins (neurotoxins) and cell-destroying toxins (cytotoxins), which can cause paralysis or severe tissue damage. These molecules often fail to trigger a strong immune response in the animals used for antibody production, meaning that the resulting treatments are only partly effective. Researchers are now exploring new approaches that bypass these challenges entirely.

“We are shifting from a discovery problem – like searching for a needle in a haystack – to a design problem, in which we just create the exact needle we need. That shift changes everything. Instead of relying on antibodies, scientists are turning to AI and advanced machine learning methods – deep learning – to create custom-designed proteins that can precisely neutralise venom toxins.”

The researchers engineer molecules that fit these toxins like a key in a lock.

“Using deep learning, we generate thousands of possible designs and then refine them to maximise their effectiveness. The result? A synthetic protein that binds tightly to neurotoxins and cytotoxins, shutting them down before they can do damage.”

Gained momentum during COVID-19

Unlike traditional antivenoms, these proteins can be manufactured using genetic engineering – recombinant DNA technology – ensuring consistent quality and reducing dependence on animal immunisation. This shift has the potential to make snakebite treatments not only more effective but also more affordable and widely available.

“Designed proteins could be a game-changer. They are smaller and more stable and can be produced in bacteria or yeast instead of horses. That means we can manufacture them anywhere, at a fraction of the cost, likely without the cold-chain nightmare that plagues traditional antivenoms,” notes Timothy Jenkins.

Traditional antivenoms must be stored and transported under strict temperature control, which makes distribution in remote areas a logistical nightmare.

“These synthetic proteins, in contrast, remain stable even at high temperatures, meaning that they can be stockpiled and transported much more easily.”

The idea of designing AI-driven proteins for snakebite treatment gained momentum during the COVID-19 pandemic, when machine learning was applied to biological problems on an unprecedented scale. A pivotal moment came when an early research article from David Baker’s laboratory at the University of Washington sparked interest among venom researchers.

“Thanks to COVID-19, I went deep into machine learning, and we started experimenting with protein design. Then, this preprint from David Baker’s group appeared, and my friend Thomas Fryer at MIT told me, ‘You should reach out to him.’ I thought, no way – he’s a giant in the field. But after a few beers, he convinced me. I sent an email, and 24 hours later, we had a meeting. That is how it all started,” remembers Timothy Jenkins.

Completely redefining it

This collaboration marked the beginning of a new chapter in antivenom research. The venom research team brought extensive expertise in snake toxins, while David Baker’s group had the computational tools needed to design synthetic proteins. Together, they started using AI-powered design tools to create novel antivenom molecules tailored to neutralise snake venom toxins more efficiently.

“It was a perfect match – Susana Vázquez Torres in David Baker’s group had already started thinking about how to apply these tools to snakebite. We had the expertise in venom research and she had the computational power. It came together like it was meant to happen,” says Timothy Jenkins.

The team did not just rely on AI to suggest potential antivenom candidates – they tested thousands of designs in real-world conditions to find the most effective ones. These tests confirmed that the designed proteins could bind to toxins with near-perfect precision.

“We saw these binders completely neutralising the toxins in functional assays – laboratory tests that measure effectiveness – but the real test was in live models. When we injected mice with a lethal dose of venom and then administered our synthetic protein treatment, every single treated mouse survived. That was the moment we knew this could be a game-changer for snakebite treatment.”

Unlike plasma-based treatments, these synthetic proteins could be stored without refrigeration, penetrate tissues more effectively and neutralise toxins before they cause irreversible damage.

“If this works, we are not just improving snakebite treatment – we are completely redefining it. Instead of harvesting antibodies from animals, we are engineering precision-made molecules that do the job better. And that could save thousands of lives every year.”

Designing proteins to neutralise snake venom

To create effective antivenom treatments, scientists needed to design proteins that could bind tightly to neurotoxins, blocking their harmful effects. Instead of relying on traditional trial-and-error approaches, they used AI to generate protein structures that fit perfectly with the toxins. By predicting how these proteins would fold, researchers could rapidly identify promising candidates before testing them in the laboratory.

“This approach bypasses many challenges of working with actual venom, making the process faster and more precise. First, we use AI to design a backbone – a structural shape matching the toxin. Then, another machine learning tool predicts the protein sequence – the blueprint for the protein. Microorganisms such as bacteria and yeast can then produce large quantities at low cost, making this approach viable for global distribution,” emphasises Timothy Jenkins.

The AI-generated designs were then filtered to select those most likely to succeed in real-world testing. Researchers synthesised the best candidates and tested them to determine whether they could block the neurotoxins effectively. Isolating specific venom components requires extensive laboratory work, including extracting toxins directly from snakes.

“Using AI-driven protein design, we could sidestep this time-consuming process and develop a highly effective binder in just weeks. We take thousands of AI-generated designs, filter out the ones most likely to work in the laboratory and then build and test them. That is how we go from theory to reality.”

Paving the way for a new generation of medical solutions

Once the most promising binders were identified, the next step was production. Instead of relying on traditional antivenom extraction, which requires immunising animals and harvesting antibodies, researchers used biotechnology to manufacture the binders in living cells. By inserting synthetic DNA into bacteria, yeast or mammalian cells, scientists could mass-produce these proteins in a controlled and scalable way – similar to how insulin is produced for medical use.

“We can use any production method we want – bacteria, yeast or mammalian cells. Novo Nordisk produces insulin using yeast, and we can do the same thing here. We synthesise the DNA sequence, insert it into an organism and let its internal machinery produce the protein. That is how we manufacture these binders – fast, scalable and without the variability of traditional antivenoms,” points out Timothy Jenkins.

This innovative approach to antivenom design not only speeds up development but also increases precision, ensuring that the resulting treatments are both effective and easy to produce. The combination of AI and biotechnology is paving the way for a new generation of medical solutions that could save lives in regions where snakebites remain a significant threat.

Neutralising venom in a controlled setting is only part of the challenge.

“In real-world snakebites, venom has already begun causing damage before treatment is administered. The next step was to determine whether these synthetic proteins could work as a post-bite rescue treatment,” highlights Timothy Jenkins.

Already solving a massive problem

In tests mimicking real snakebite scenarios, researchers administered the binders after the venom had already begun acting. The results were astonishing: even when given with a delay, the designed proteins rapidly neutralised the toxins, preventing fatal effects. If this is proven safe, this alongside the proteins being so stable opens the door to new, more accessible treatments that could be self-administered in emergencies, potentially in an EpiPen-like format.

“We did not just test whether the binders worked when mixed with venom beforehand – that is not how real snakebites happen. We ran rescue assays: injecting venom first, then administering the binders with a delay. The results? Even at very low doses, our binders neutralised the toxins. And we are talking about concentrations that are likely orders of magnitude lower than existing treatments, making this far more practical for real-world use,” explains Timothy Jenkins.

Beyond their effectiveness, these synthetic proteins have key advantages over traditional treatments. Their stability, easy production and precision reduce risks such as allergic reactions and high doses. AI has enabled a more predictable, scalable solution, paving the way for next-generation antivenoms that are safer and more rapid and save lives in remote areas with limited hospital access.

“If we create better injectables, we are already solving a major problem. Current treatments require intravenous infusions in hospitals, but these designed proteins are stable, easy to produce and work as intended. This could lead to an EpiPen-style emergency treatment – rapid, simple and life-saving. Traditional antibodies need large doses and pose risks such as anaphylaxis, but our binders are precise, reproducible and likely a lot safer.”

Changing everything

Antivenoms have long been the gold standard for treating snakebites, but they have significant limitations. The process of creating these treatments involves immunising animals, which not only makes production expensive but also leads to variation in effectiveness and potential adverse reactions. Unlike traditional antivenoms, they are designed to target specific toxins with greater accuracy, potentially revolutionising snakebite treatment.

“We have proven that this works. The next step is making it practical – building a complete cocktail of binders tailored to specific regions, such as India or sub-Saharan Africa. We know that the technology is solid, but now comes the real work: engineering a treatment that covers multiple species and venom types. Thankfully, I just secured a grant to make this happen,” remarks Timothy Jenkins.

Over the next three years, the researchers will be scaling up – turning this from an experimental breakthrough into a real-world solution.

“The ability to design targeted synthetic binders does not just change snakebite treatment – it could revolutionise medicine as a whole. Researchers are already exploring how these same AI-driven techniques can be used for other conditions, from cancer to infectious diseases.”

The impact of AI-designed proteins goes beyond snakebites. Timothy Jenkins and his team are using similar methods to develop targeted cancer treatments, leveraging machine learning to create precise therapies. The same approach that neutralises venom helps the immune system to attack cancer cells. This precision is a game-changer, enabling researchers to predict how binders attach to targets, boosting effectiveness while reducing side-effects.

“AI-designed proteins are going to change everything – diagnostics, reagents and therapeutics. We have already demonstrated we can design binders that target cancer cells for chimeric antigen receptor T-cell therapy. That is the same core technology we are using to neutralise snake venom. It is all about precision design, and the implications go far beyond medicine. This is not just an improvement; it is a paradigm shift.”

Potential to democratise drug discovery

Turning these scientific breakthroughs into accessible treatments presents challenges. Although the technology is advancing rapidly, regulatory approval remains a hurdle. AI-designed proteins are new, and agencies are cautious with non-traditional treatments. Researchers are working with global health organisations to navigate the approval process and ensure that these solutions reach those who need them most.

“The technical challenges? We will solve them. The regulatory challenges? Those are the real beast. We are chipping away at it, engaging with global health organisations, working through the requirements for clinical trials. The good news is that the snakebite research community is coming together on this, and WHO has already outlined target product profiles. The pieces are there – we just need to push them into place,” says Timothy Jenkins.

Beyond regulation, another crucial aspect is ensuring that these innovations benefit the regions most affected by snakebites. AI-driven research has the potential to democratise drug discovery, enabling scientists from lower-income countries to contribute without needing high-tech laboratories. This could mark a shift in how treatments for neglected diseases are developed, making life-saving medicines more accessible and affordable for the people who need them the most.

“Generative AI is opening doors for new talent. With just an Internet connection and a basic computer, brilliant minds anywhere can drive innovation. This is crucial for neglected diseases, in which those most affected have had little chance to contribute. If done right, AI-driven research could democratise healthcare, making drug discovery more accessible and fairer.”

The future of antivenom development is changing rapidly. From traditional animal-derived antibodies to cutting-edge AI-designed proteins, the landscape is evolving at an unprecedented pace. What was once a field relying on labour-intensive and expensive methods is now entering an era of rapid, scalable solutions driven by computational power.

“The potential is enormous – not just for snakebites but for a wide range of medical challenges. With the right support and regulatory pathways, AI-designed proteins could redefine how we tackle some of the world’s most persistent health problems,” concludes Timothy Jenkins.