Venom is a potent mix of molecules evolved to paralyze, kill, and destroy cells. But it could also hold some promising tools for the future.
In a new Nature Communications study, researchers at the University of Pennsylvania used artificial intelligence to analyze millions of compounds from snake, scorpion, and spider venom. They identified promising candidates for new antibiotics.
“Venoms are evolutionary masterpieces,” said César de la Fuente, a synthetic biologist at Penn. “They’ve spent hundreds of millions of years learning how to breach diverse biological defenses.” His team built an algorithm to predict which venom peptides could fight drug-resistant bacteria.
Algorithms Take Aim at Superbugs
Before antibiotics, even minor infections like a cut on the hand or a sore throat could turn deadly. These drugs transformed medicine, making surgeries safer, childbirth less risky, and once-lethal diseases treatable. But bacteria are evolving faster than we can develop new drugs, and many are now resistant to multiple antibiotics. This means infections we thought we had conquered are making a comeback, threatening to undo a century of medical progress and putting millions of lives at risk worldwide.
The study centers around a deep-learning tool named APEX, designed to predict antimicrobial activity from protein sequences. Trained on known antibiotics, APEX was deployed to mine 16,000 venom proteins sourced from snakes, spiders, scorpions, cone snails, and even sea anemones.
Then, using a sliding window approach, the team generated a staggering 40 million short peptide sequences called venom-encrypted peptides, or VEPs. These were scored based on how likely they were to act like antibiotics. Within hours, the algorithm narrowed the list down to 386 strong candidates.
“APEX lets us scan an immense chemical space in just hours and identify peptides with exceptional potential,” said Marcelo Torres, a co-author on the study.
From there, the researchers chemically synthesized 58 of the most promising VEPs and tested them against 11 different pathogens, including Acinetobacter baumannii and Pseudomonas aeruginosa—bacteria notorious for their antibiotic resistance.
The results were stunning. Of the 58 VEPs tested, 53 showed potent antibacterial activity.
Peptide Testdrive
To assess whether the peptides could work in real-life conditions, the team tested three of them in mice infected with A. baumannii, a bacteria notorious for surviving in hospitals and resisting treatment.
After infecting the animals with a skin abscess, they treated the area with a single dose of each VEP. Two days later, bacterial counts had dropped sharply. One of the peptides, derived from the venom of the wolf spider Geolycosa riograndae, reduced bacterial load by 99.9%—a result on par with standard antibiotics.
The mice showed no signs of weight loss or toxicity, suggesting the VEPs were not only effective but also safe in this model.
The motivation behind this study is deadly serious. Drug-resistant infections are responsible for an estimated 5 million deaths each year, according to global health estimates. And the antibiotic discovery pipeline has all but dried up.
“Despite the growing threat, the development pipeline for novel antibiotics has stagnated over the past few decades due to high costs and lengthy timelines,” the study authors wrote. In response, scientists are turning to overlooked sources—like animal venom.
Historically, venom has already yielded life-changing medicines. The painkiller ziconotide was derived from cone snail venom. The blood pressure drug captopril came from snake venom. Even semaglutide, the popular diabetes and weight-loss drug, was inspired by a hormone found in the Gila monster’s saliva.
But using venom to fight bacteria is rather novel.
“Venom compounds are fast acting, very potent, and very specific,” said Mandë Holford, a chemical biologist at Hunter College who was not involved in the study, as per The Scientist. “All the ingredients you look for when you’re trying to make a drug.”
What’s Next?
Though promising, the research is in its early stages. The team is now optimizing the top candidates through medicinal chemistry—tweaking their structure to make them more stable, less toxic, and longer lasting. There are plenty of substances that hold antibiotic promise but never made it to treatment.
The scientists behind this work also acknowledge limitations. For instance, the APEX model currently works only on peptides under 50 amino acids and doesn’t yet explain why certain sequences work better than others. But that may change.
Introducing features like self-attention mechanisms or integrating large language models could help explain why a given venom fragment is so deadly to bacteria, and perhaps even point the way to safer and more effective antibiotics.
Antibiotics were once considered a miracle. But overuse has allowed microbes to adapt and resist. Bacteria can now survive drugs that once worked reliably—an evolutionary arms race that humans are losing.
What this new study comes in as a plot twist. By turning to the very molecules animals evolved to kill, scientists might discover how to save lives. With artificial intelligence doing the heavy lifting, that discovery process may be faster than ever.
