The search for new antibiotics is leading scientists to explore unusual sources, including artificial intelligence and the genetic material of extinct species. Bioengineer César de la Fuente and his team have been studying a compound called mammuthusin, which has demonstrated the ability to eradicate superbugs in laboratory settings. Superbugs are bacteria that have developed resistance to existing antibiotics, making infections tough to treat.

According to de la Fuente, a professor at the University of Pennsylvania, the increasing ineffectiveness of current antibiotics has created an urgent need for novel solutions. He and other researchers are investigating sources ranging from Neanderthal DNA to ordinary soil in the hopes of discovering new antibacterial agents. “Antibiotic resistance is one of the greatest challenges we face as a society,” de la fuente stated.

Superbug infections contribute to over five million deaths worldwide each year, and this number is on the rise. Common infections are becoming increasingly difficult to treat as antibiotics lose their effectiveness. Without the development of new drugs, antibiotic resistance could lead to approximately 39 million deaths by 2050, according to a recent study.

The pharmaceutical industry has been slow to develop new antibiotics, hampered by the rapid evolution of bacterial defenses. Historically, many of the antibiotics that have saved countless lives were derived from natural sources, often discovered by chance. A prime example is Alexander Fleming‘s accidental finding of penicillin in 1928.

To effectively combat superbugs, medical professionals emphasize the need for antibiotics with unique chemical structures and mechanisms. However, only a limited number of such drugs have been introduced to the market in recent decades.

AI-Powered Discovery

De la Fuente is optimistic that artificial intelligence can definitely help accelerate the discovery of new antibiotics.His team has developed deep-learning algorithms to analyze extensive genetic databases, identifying peptides with antibacterial properties. This approach has been used to study animal venoms, the human microbiome, archaea, and even the genetic codes of Neanderthals and Denisovans. De la Fuente notes that their “deep-learning model has opened a window into the past.”

“Antibiotic resistance is one of the greatest challenges we face as a society.”

Conventional antibiotics are typically small-molecule drugs derived from bacteria and fungi,easily administered as pills due to their ability to penetrate cell membranes. Peptides, composed of short chains of amino acids, are larger and more complex, making them less stable and harder to formulate into pills.

Recent advancements have improved the absorption and usability of peptide drugs, including certain IV antibiotics and insulin. Antibacterial peptides are abundant in nature, serving as part of the immune system in many organisms.

De la Fuente believes that “peptides are the next big thing in medicine.” In January, he launched a startup to further explore the potential of mammuthusin and other peptides as antibiotics.

When the algorithms identify promising peptides, de la Fuente’s team uses robots to synthesize the compounds and test their effectiveness in mice infected with bacteria.Initial results have been promising, with several hundred peptides demonstrating the ability to safely and effectively cure infections in mice.

Reviving Ancient defenses

One such peptide is mammuthusin, identified from the genetic code of Mammuthus primigenius, the woolly mammoth. Researchers discovered it by analyzing DNA sequencing data from fossil remains in a National Center for Biotechnology Information database. Mammuthusin proved as effective as polymyxin B, a last-resort antibiotic, in eradicating a bacterium designated as a critical pathogen by the World Health Association due to its resistance to multiple antibiotics. These findings were published in the journal Nature.

James Collins, a bioengineer at the Massachusetts Institute of Technology, emphasizes that working with extinct species “is expanding the chemical space that we could explore,” offering access to molecules that evolved in different environments and time periods.

Collins’s lab is also using algorithms to search chemical databases for potential antibacterial compounds and experimenting with generative AI to design entirely new molecules. In a 2024 paper, Collins and his colleagues reported identifying structurally unique antibiotics after analyzing over 12 million chemical compounds.

Gerry Wright, a biochemist at McMaster University in Canada, suggests that modern scientific techniques can reveal new opportunities in previously explored natural environments. He notes that historically,antibiotics were often discovered by examining soil samples for fungi and bacteria. Microbes have developed elegant defense mechanisms through eons of competition. “There haven’t been any compounds better than those made in nature,” Wright says.

However, by the 1990s, researchers exploring natural environments for antibiotics encountered a hurdle: the repeated rediscovery of the same compounds.This led to a sense of frustration and the misconception that natural sources had been tired, according to Wright.

wright’s lab has overcome this challenge by using genetic sequencing and chromatography to analyze known microbes and identify previously overlooked antibacterial molecules.

Persistence has also been key.One researcher in Wright’s lab incubated a soil sample from a backyard in Ontario for an entire year, without adding nutrients, to encourage the growth of rarer organisms.

This approach led to the discovery of a slow-growing species that produced both a common antibiotic and a novel antibacterial peptide. This peptide effectively eradicated drug-resistant bacteria in mice, as reported in a March paper in Nature.

Wright concludes, “We are going back with different eyes and different techniques and discovering that there’s still a lot of wealth to be found.”