For decades, bacteria have been waging microscopic battles against viruses, wielding molecular weapons that scientists barely understood—until now. Two groundbreaking studies published in Science reveal the vast, previously hidden diversity of bacterial immune systems. Using advanced machine-learning algorithms to scan thousands of bacterial genomes, the research teams identified hundreds of thousands of potential antiviral proteins, a treasure trove that could inspire a new era of molecular engineering.
“This is a treasure trove for any biochemist,” said José Antonio Escudero, a microbiologist at the CSIC National Center for Biotechnology in Madrid, who was not involved in either study. The discoveries extend the legacy of CRISPR–Cas9 and restriction enzymes—two bacterial immune systems that were repurposed as revolutionary genetic tools. Scientists hope these newly identified proteins may similarly yield the next generation of molecular technologies.
Previous work had confirmed that bacteria deploy more than 250 proteins to defend against viral infections, but researchers suspected the true diversity of bacterial immunity was far greater. “The big question was how much diversity and how can we actually predict it at scale?” said Aude Bernheim, a microbiologist at the Pasteur Institute in Paris and co-author of one of the studies. Using deep-learning models trained on protein and genomic data, Bernheim’s team predicted that on average 1.5% of genes in a bacterial genome contribute to antiviral immunity—three times higher than earlier estimates. Remarkably, more than 85% of the predicted protein families had never before been linked to immune functions. Laboratory experiments confirmed 12 previously unknown “antiphage” systems in Escherichia coli and Streptomyces albus, proving the predictive models were uncovering genuine biological defenses.
In a complementary approach, Michael Laub and colleagues at the Massachusetts Institute of Technology created a machine-learning tool called DefensePredictor. Trained on gene and protein data from 17,000 bacterial genomes, the tool identified 624 defense-related proteins in 69 strains of E. coli, including over 100 previously uncharacterized immune proteins. Laboratory testing validated the antiviral activity in 42 of these cases, demonstrating that the algorithm can accurately pinpoint functional bacterial immune systems.
Both studies converge on the same conclusion: scientists have been dramatically underestimating the number of bacterial defense systems. “This research brings to light how many systems are out there still to be characterized,” said Laub, highlighting the potential for these proteins to become powerful tools for biotechnology. Escudero added, “The discoveries include hundreds of genes that we didn’t know were related to immunity.”
To facilitate future research, both teams have made their findings openly accessible. Laub’s DefensePredictor is available online for researchers to explore, while Bernheim’s team created DefenseFinder, a database containing over 44,000 predicted antiviral systems. These resources allow scientists to test the properties of newly discovered proteins and investigate the molecular mechanisms through which bacterial immune systems detect and neutralize viral infections. “There are really interesting questions about how they sense phage infection and how they block a phage from replicating,” Laub noted.
The implications extend far beyond microbiology. These newly identified systems could serve as templates for antiviral drugs or precise molecular tools, potentially sparking a “biotechnological revolution” comparable to CRISPR, according to Escudero. They also offer insights into the evolutionary history of immunity, revealing that mechanisms now central to mammalian immune systems may have ancient bacterial origins. Bernheim emphasized, “This really pushes what the field can bring—not only to microbiology, but also to a much broader picture of what immunity is across domains of life.”
By mining the hidden “dark matter” of bacterial genomes, researchers are uncovering a staggering arsenal of molecular defenses that has evolved over billions of years. These proteins, once obscured in the genetic code, may now form the foundation for transformative technologies in medicine, synthetic biology, and genetic engineering. The studies illustrate how cutting-edge computational tools—deep learning, large-scale genomic analysis, and predictive modeling—are accelerating the discovery of nature’s most sophisticated molecular strategies.
As scientists continue to explore this treasure trove, the line between natural immunity and engineered molecular tools may blur, ushering in a new era where bacterial strategies against viruses inspire human-designed technologies. The next CRISPR-like breakthrough may already exist in the genomes of bacteria, waiting to be discovered.
