Bites and stings from spiders, snakes, and scorpions are usually painful and at times dangerous. Researchers at the University of Pennsylvania, however, see these animal venoms as a rich and largely untapped source of bioactive molecules that may include antimicrobial agents. This approach is timely: antibiotic‑resistant infections, particularly those caused by Gram‑negative bacteria, are rising, and experts estimate that drug‑resistant infections contribute to roughly 5 million deaths per year while traditional antibiotic discovery has stalled.
Venoms as evolutionary libraries of bioactive peptides
The team points out that venoms are evolutionary products refined over hundreds of millions of years to overcome biological defenses. Venoms contain diverse bioactive peptides and proteins with multiple pharmacological effects, including antibacterial activity. Building on that idea, the researchers developed an artificial intelligence pipeline to predict which venom peptides could overcome drug‑resistant bacteria.
APEX deep learning model and large‑scale peptide generation
They created a deep‑learning tool called APEX to predict antimicrobial activity from protein sequences. APEX was trained on known antibiotics and then applied to roughly 16,000 venom proteins from snakes, spiders, scorpions, cone snails, and even sea anemones. Using a sliding‑window approach, the team generated more than 40.6 million short peptide sequences derived from those proteins; they termed these venom‑encoded peptides (VEPs).
APEX scored the VEPs by the likelihood they would behave like antibiotics. Within hours the algorithm narrowed candidates down to 386 strong hits. The researchers chemically synthesized the 58 most promising VEPs and tested them against 11 different pathogens, including some of the most notorious antibiotic‑resistant bacteria. Of the 58 tested compounds, 53 exhibited strong antibacterial activity.
Mouse experiments against multidrug‑resistant Acinetobacter baumannii
The team then evaluated three VEPs in mice. The animals were infected with Acinetobacter baumannii, a hospital‑acquired pathogen known for resistance to multiple antibiotics. A single dose of VEP treatment produced a steep decline in bacterial counts. One peptide derived from wolf spider venom reduced bacterial burden by 99.9 percent—an outcome comparable to standard antibiotics. The treated mice did not show notable adverse effects, suggesting the intervention was both effective and tolerable in these preclinical tests.
Next steps and cautionary note
While the results are promising, the work remains at an early stage. The lead peptides are being optimized via medicinal chemistry to improve stability, reduce toxicity, and extend duration of action. The authors also caution that many compounds with preclinical promise never become approved drugs, so further development and clinical testing will be required to determine whether these VEPs can translate into safe, effective therapies.
The study was published in Nature Communications and illustrates how combining natural product diversity with artificial intelligence can accelerate the search for new antimicrobial agents.
