Their algorithm studied the entire human proteome and came up with a rough list of about 43,000 peptides. Torres narrows it down to 2,603 that come from proteins known to be secreted from cells. Some are complete small proteins and hormones. Everything else is just fragments, strings encoded in a much larger complex. None of them have been previously described as antibiotics.
To check if their AI is on the right track, Torres compiled 55 of the most promising candidates. He tested each one in liquid samples against the “who is who” of resistant bacteria: Pseudomonas aeruginosa, a well-known failing organ of the lung; Acinetobacter baumannii, known to be rampant in hospitals; Staphylococcus aureus, the germ that causes dangerous staph infections — plus others, for a total of eight. Of the 55 people, most were able to stop the bacteria from multiplying.
Several peptides stand out, including SCUB1-SKE25 and SCUB3-MLP22. These peptides live along regions known as “CUB domains” that exist in proteins involved in a long list of functions such as fertilization, creating new blood vessels and suppressing tumors. SCUBs are just parts of the whole. But in essence, they seem to be very adept at killing germs. So Torres pushed these two SCUBs to test in mice.
Torres tested whether SCUB, or a combination of the two, could clear the infection in mice with infections under the skin or in their thigh muscles (a model for more systemic disease). In all cases, the bacterial populations sampled from these tissues stopped growing. And in some cases, as Torres noticed on his warm agar, the bacterial count plummeted.
Torres also examined how easily bacteria can evolve resistance to the peptides, compared with an existing antibiotic called polymyxin B. After 30 days of exposure, bacteria can tolerate high doses of polymyxin B 256 times more than the original amount, but SCUBs are still effective at the same dose. (It takes a lot of genetic change for bacteria to adapt to membrane damage.) Of course, that doesn’t mean they will never adapt, especially over longer periods of time. “Nothing can resist resistance,” says De la Fuente. “Because bacteria are the most evolved species that we know of.”
As systematic as the team’s plan was, Torres was still a bit stunned. “We think we are going to have a lot of success,” he said of the peptides revealed by the AI. But to his surprise, the peptides came from all over the body. They are from proteins in the eyes, nervous and cardiovascular systems, not just the immune system. “They’re literally everywhere,” says Torres.
The team suggests that life evolved this way to pack as much punch into the genome as possible. “A gene codes for a protein, but that protein has many functions,” says de la Fuente. “I think this is really a clever way to evolve keeping genomic information to a minimum.”
This is the first time that scientists have found antibiotic peptides in proteins that are not involved in the immune response. Jon Stokes, a biochemist at McMaster University in Canada who was not involved in the study, said the idea was “really innovative, but prepared his lab to incorporate AI in finding search for small molecule antibiotics. “The take-home approach for me was: Start looking for antibiotics in hidden places.”
