Their algorithm disrupted all human proteomes and spewed out a series of about 43,000 peptides. Torres reduced the amount to 2,603 derived from proteins known to be derived from cells. Some were small amounts of protein and hormones. Some were pieces, chains hidden inside a very large loop. No one was ever reported to be on antibiotics.
To see if their AI was on the right track, Torres made 55 of the most promising. He tested everyone in fluid samples against the “who” of antibiotics: Pseudomonas aeruginosa, the most common lung disease; Acinetobacter baumannii, which is known to be widespread in hospitals; Staphylococcus aureus, staph virus infection — plus some eight. At 55, many were able to prevent the bacteria from recurring.
A few peptides were identified, including SCUB1-SKE25 and SCUB3-MLP22. Peptides live in so-called “CUB domains” that are found in proteins that have a long range of functions such as semen, making new blood vessels, and suppressing tumors. SCUBs are complete pieces. But on their own, they seemed to excel at killing germs. That’s why Torres recommended the two SCUBs to be tested in mice.
Torres tested whether SCUB, or a combination of the two, could alleviate the disease of infected rats under their skin, or their thigh muscles (an example of a more common disease). Invariably, the amount of bacteria extracted from these tissues stopped growing. And sometimes, as Torres noticed on his warm agar, the amount of bacteria dropped.
Torres also tested how bacteria can easily convert peptides, compared to an existing drug called polymyxin B. After 30 days of exposure, the bacteria can tolerate polymyxin B levels 256 times higher than their original volume, but SCUBs. remained active at the same level. (Genetic mutations are needed for bacteria to adapt to membrane damage.) But this does not mean that they cannot change, especially over time. “Nothing can be further from the truth,” says de la Fuente. “Because bacteria are the most mutated species we know.”
As the team’s plans were in place, Torres was still left out. “We thought we would have a lot of beats,” he says of the peptides revealed by AI. But surprisingly, peptides came from all over the body. It was based on eye proteins, blood vessels, and the cardiovascular system, not just the immune system. “She is everywhere,” says Torres.
The team thinks that life evolved in this way to get as many punches as possible into the genome. “One gene contains one protein, but that protein has many functions,” de de Fuente said. “This is a realistic, I-logical, evolutionary approach to reducing genomic consciousness.”
This is the first time that scientists have found antibiotic peptides inside proteins that are not immune to the immune response. The idea was “really creative,” says Jon Stokes, a medical scientist at the University of McMaster, Canada, who did not participate in the study, but has been preparing his lab to incorporate AI into microbial research. “Here’s what I take home with me: Start looking in unfamiliar places to find antibiotics.”