Stanford AI Discovers Bacteria-Killing Polymers as a New Class of Antibiotics

Conventional pharmaceutical development assumes that an effective antibiotic must infiltrate a bacterium and disable a specific internal protein—a precision that allows superbugs to quickly evolve resistance by altering that single target. Now, a Stanford University engineering team has demonstrated a radically different approach, using artificial intelligence to discover a new class of synthetic polymers that simply rip holes in the bacterial membrane from the outside.[2][4]

The mechanism is borrowed from antimicrobial peptides, which are small proteins naturally present in human blood and sweat. These peptides kill bacteria on contact by physically permeabilizing the cell membrane, a brute-force structural attack that microbes struggle to evolve defenses against. However, natural peptides are notoriously difficult to adapt into mass-produced drugs because they degrade rapidly and are expensive to synthesize at scale.[2][3]

To solve this, materials science professor Eric Appel and postdoctoral scholar Shoshana Williams sought to engineer synthetic polyacrylamides—long, chain-like molecules—that mimic the physical attack of peptides but remain stable and cheap to produce. "Antimicrobial peptides are chemically able to get very close to and disrupt the cell membrane, killing the bacteria," Williams explained. "Importantly, they don't need to get inside the cell to work, like a typical drug would. Nor do they work on one specific protein or pathway, like drugs do."[2][3]

Finding the right polymer required sifting through a computationally generated library of 1.7 million potential structures. The researchers immediately encountered a common bottleneck in medical artificial intelligence: a severe lack of experimental data. While extensive datasets exist detailing the properties of antimicrobial peptides, data on antimicrobial polymers is scarce, leaving the AI without enough examples to learn from.[2][4]

The Stanford team bypassed this limitation using a technique called cross-molecular active learning. They first trained their predictive algorithms on the abundant peptide data, teaching the model the chemical traits that allow a molecule to rupture a membrane. They then applied that framework to the 1.7 million polymer candidates. To rapidly refine the model's accuracy, the researchers synthesized and tested the 20 specific polymer candidates where the algorithms disagreed the most, feeding the results back into the system.[2][4]

This iterative feedback loop allowed the AI to isolate 10 highly potent antimicrobial leads. In laboratory tests published September 21, 2026, in the journal Matter, all 10 candidates exceeded performance benchmarks against Escherichia coli (E. coli). The polymers successfully targeted Gram-negative bacteria, a class of double-membraned microbes that includes Salmonella and for which no new class of antibiotics has been introduced in decades.[2][4]

One specific copolymer demonstrated exceptional efficacy against biofilms, the dense, protective microbial communities that routinely withstand conventional antibiotic treatments. When researchers combined this copolymer with a standard clinical drug regimen, it reduced the amount of the traditional antibiotic needed to eradicate biofilm-associated E. coli by three orders of magnitude—a 1,000-fold reduction.[4]

This shift toward materials science and physical disruption is gaining traction across the broader field of infection control. On September 25, 2026, a separate research team from Wichita State University and the University of Kansas School of Medicine reported the development of selenium-enriched acetic acid and gelatin hydrogels. Designed specifically for burn injuries, which affect more than 11 million people annually and are highly vulnerable to superbug infections, these hydrogels maintain 90 percent moisture while providing intrinsic antimicrobial protection without relying on traditional biochemical drugs.[1][5]

For the Stanford team, the immediate next hurdle involves animal modeling to confirm that these membrane-disrupting polymers can retain their bacterial lethality without damaging healthy human tissue. If the safety profile holds, the computational pipeline could be adapted to design custom polymers for specific pathogens. "We really do need better drugs," Appel said. "This new process opens a promising path to identifying novel antibiotics that work in new and different ways to treat serious and complicated infections and combat resistance."[2][4]