AI Finds Phage Mutations That Made Bacteria-Infecting Viruses Up to 1 Million Times More Effective
phage
A virus that infects and destroys particular bacteria.
antibiotic resistance
A condition in which antibiotics work less well against bacteria.
mutation
A small change in a living thing's instructions.
What happened
Researchers at the University of Wisconsin–Madison built an AI model for designing phage mutations. Phages are viruses that infect and destroy particular bacteria. The team used laboratory data to teach the model which changes might help a phage recognize and infect a bacterial host.
The researchers then engineered phages with mutations suggested by the model. They tested those phages in the laboratory. Some infected their bacterial hosts up to six orders of magnitude more effectively than natural phages. Six orders of magnitude means up to 1,000,000 times in this comparison. It does not mean a million times more healing for a patient.
The problem behind the work
Antibiotics have long been used against bacterial illnesses. But bacteria can evolve antibiotic resistance, making those medicines less effective. That leaves fewer reliable treatments for some infections.
Phage therapy offers another possible approach. It uses natural or engineered phages to infect and destroy selected bacteria. This approach has its own challenge. A natural phage does not always evolve for maximum killing power. If it eliminates every available host, it loses the place where it reproduces. Bacteria also evolve defenses against phage infection.
Why the result matters
The Wisconsin team is not searching only for stronger phages. It is also searching for more precise ones. The model was trained to identify mutations that could target selected bacteria while sparing helpful bacteria in the gut.
That distinction matters. A useful treatment may need to attack a disease-causing bacterium without disturbing other microbes. The model could also shorten the search for promising candidates. The research group has measured how tens of thousands of mutations affect phage proteins. AI can learn patterns from that map and suggest possibilities that researchers might otherwise test much more slowly.
What the study actually showed
The model rests on years of high-throughput experiments in the Raman laboratory. In one line of work, the researchers studied a phage protein that acts like a molecular drill. It helps the phage pass through a protective sugar coating around some bacteria. The experiments showed how amino-acid changes affected bacterial recognition and infection.
Data from another phage protein helped form the foundation for the new model. The team tested model suggestions by engineering phages and measuring their performance in the laboratory. The reported million-fold result comes from that laboratory comparison.
The researchers say they are moving toward pathogens linked to urinary tract infections, microbiome-related diseases, and Salmonella in poultry. These are development directions. They are not approved treatments.
What remains unknown
The report does not present patient treatment results or results inside a human body. It does not establish that every AI suggestion will work. It also does not show whether the same performance will hold across different bacterial strains and real infection conditions.
The million-fold figure describes infection efficiency against a bacterial host. It is not a cure rate, a survival rate, or a measure of safety. The model's ability to spare helpful bacteria also needs testing beyond the reported experiments.
What to watch next
The next important evidence will come from repeated tests outside the original laboratory system. Researchers will need to see whether the phages keep their target precision and useful activity in more realistic settings.
The team has named urinary infections, microbiome-linked diseases, and poultry Salmonella as possible targets. Progress in those areas will show whether AI-designed mutation candidates can become practical phage treatments. For now, the study shows a faster way to search. It does not yet show a finished medicine.
Source: Phys.org report, September 24, 2026.
AI Helps Build Viruses That Attack Specific Bacteria
📰 Full story: AI Finds Phage Mutations That Made Bacteria-Infecting Viruses Up to 1 Million Times More Effective
Researchers used AI to find phage changes that worked better in lab tests.
phage
A virus that attacks specific bacteria.
antibiotic resistance
When antibiotics no longer kill bacteria well.
mutation
A small change in a phage's instructions.
💡 The gist
- AI helped researchers choose changes for bacteria-fighting viruses.
- Some changed phages infected bacteria up to 1,000,000 times better.
- The result is a lab finding, not a patient treatment.
The University of Wisconsin–Madison, a research university, built the AI model. It studied phages. Phages are viruses that attack specific bacteria.
Antibiotics are medicines that fight bacterial illnesses. Some bacteria develop antibiotic resistance. Then the medicine works less well. Doctors and scientists need other ways to fight those infections.
Phage therapy is one possible way. It uses phages to infect and destroy selected bacteria. Natural phages have limits. If they destroy every host, they lose the place where they reproduce. Bacteria can also build defenses against them.
Researchers fed the AI model data from many laboratory experiments. Those experiments changed phage proteins and measured infection. Some proteins help phages recognize bacterial coatings. Other changes affect how well phages infect their hosts.
The model suggested new mutations. A mutation is a small change in a phage's instructions. Researchers built phages with the suggested mutations. Then they tested those phages against bacteria.
Some phages infected bacteria up to 1,000,000 times more effectively. That number describes infection efficiency in the laboratory. It does not mean people can be cured 1,000,000 times better.
The model also searches for more specific phages. The goal is to attack selected bacteria while sparing helpful gut bacteria. This could matter for future treatments.
The team is exploring bacteria linked to urinary infections, microbiome-related diseases, and poultry Salmonella. These are research directions. No patient treatment was reported.
Scientists still need more tests. They must check whether the phages work in realistic infections. They must also check safety and unwanted effects.
Source: Phys.org report, September 24, 2026.
A Computer Helper Finds Stronger Bacteria-Fighting Viruses
📰 Full story: AI Finds Phage Mutations That Made Bacteria-Infecting Viruses Up to 1 Million Times More Effective
Scientists found changes that helped special viruses attack bacteria.
phage
A virus that attacks bacteria.
antibiotics
Medicines that fight bacteria.
Some bacteria can make people sick. Antibiotics are medicines for these illnesses. Sometimes bacteria stop responding to them.
Scientists at the University of Wisconsin–Madison, a research university, studied phages. Phages are viruses that attack bacteria.
The scientists gave an AI model many lab results. The model looked for helpful tiny changes. Scientists made phages with those changes.
In tests, some phages infected bacteria up to 1,000,000 times better. This number describes the laboratory test. It does not mean sick people are cured.
Scientists have not reported patient treatment. More tests must come next. Researchers want to attack harmful bacteria. They also want to spare helpful gut bacteria.
Source: Phys.org report, September 24, 2026.