Scientists at the University of Pennsylvania have used artificial intelligence to uncover antibiotics hidden in venom. And yes, we’re talking about the kind found in snakes, spiders, and scorpions. The research team, led by bioengineer César de la Fuente, fed a machine-learning model data from over 40 million peptide fragments taken from venom proteins, aiming to identify molecules with antibiotic potential. The results, published in Cell, mark one of the most promising developments in tackling antibiotic resistance in recent years.
AI sifted through venom from thousands of animals, including cone snails and scorpions, and narrowed the possibilities to 386 candidates that showed potential to fight bacteria. Out of those, the researchers synthesised 58 and tested them in the lab. An impressive 53 of them, about 91%, were effective at killing bacteria without harming human red blood cells. In other words, they hit the sweet spot: deadly to bugs, harmless to us.
A venomous solution to antibiotic resistance? Yes, please!
One of the standout peptides came from a spider. When tested on mice infected with Acinetobacter baumannii, a multidrug-resistant bacterium classified by the World Health Organization as a “critical priority pathogen,” a single topical dose reduced the infection by 90% with no toxicity. According to Phys.org, the mouse model success adds serious weight to the idea that venom peptides could become viable treatments in humans.
What makes these peptides particularly valuable is how they attack bacteria. Most antibiotics work by targeting specific cellular machinery, which bacteria can often outmanoeuvre through mutations. But these venom-derived compounds take a more brute-force route: they punch holes in bacterial membranes. That mechanism makes it much harder for bacteria to develop resistance. As antimicrobial resistance continues to rise, with some estimates tying it to more than five million deaths globally each year, this approach could be a game-changer.
And this isn’t just about making new drugs from venom for novelty’s sake. The peptides’ behaviour was studied at scale, and researchers found that 26 of them worked by collapsing the bacteria’s membrane potential, essentially knocking out the bacteria’s ability to control its own internal environment. The research also uncovered over 2,000 novel antimicrobial motifs, short molecular patterns that can be used to guide the design of next-generation antibiotics. These were revealed using further machine learning steps, showing the power of AI not just to select potential compounds, but to generate entirely new ideas for drug design.
AI speeds up what nature already perfected.
One of the main reasons this research matters so much is the sheer speed of the discovery process. Traditional antibiotic discovery takes years and often involves screening thousands of candidates manually in the lab. With the AI model APEX, short for “Automated Peptide Exploration,” the researchers could process tens of millions of possibilities in hours. It’s a massive acceleration of what used to be a painfully slow pipeline.
And they’re not stopping here. The team is now working to tweak the most promising peptides to improve their stability and longevity in the body. Many naturally occurring peptides degrade quickly in the bloodstream, so the researchers are experimenting with chemical modifications like terminal capping and the introduction of non-natural amino acids to keep them stable for longer. This work is about building practical, deliverable treatments, not just interesting lab results.
This research also proves that venom, long thought of as solely dangerous, is more like a complex biochemical toolkit. It’s evolved over millions of years to target nerves, blood, and cellular functions with extreme precision. And thanks to AI, scientists can now map and repurpose that precision. As Phys.org reported, this could open up entirely new frontiers, not only in antibiotics, but in antiviral and antifungal medicines too.
What happens next?
It’s early days, of course. These peptides still need to go through full clinical trials, and regulatory hurdles remain. But given the increasing urgency around antibiotic resistance, particularly in hospitals where infections like MRSA and carbapenem-resistant bacteria are becoming harder to treat, venom-based drugs might arrive sooner than you’d think, especially for topical applications like wound treatments, where approval tends to be quicker.
There’s also the cost to consider. Developing new antibiotics is expensive and doesn’t always offer the commercial returns that large pharmaceutical companies want. That’s why many labs, including de la Fuente’s, are working on open-access platforms and collaborations to get these discoveries moving faster. AI makes it easier to test huge chemical libraries on the cheap, levelling the playing field between universities and big pharma.
If these compounds make it through further testing, we could be looking at a whole new way to treat infections. Venom, once something we feared, could end up being what saves us from one of the greatest health threats of the modern age.