AI-Found Molecule Cut Food Intake by 50% in Animals

Treated animals also showed improved glucose and insulin tolerance — measures of how well the body handles sugar.

How did AI find it?

The body produces enormous numbers of small protein fragments, and only a rare few act as hormones. Many hormones start as large inactive proteins called prohormones, which enzymes slice into smaller peptides. A single prohormone can be cut in several places, producing many candidate fragments, most of them biological leftovers.

Traditional searches extract proteins from tissue and analyze them with mass spectrometry — a method that identifies molecules by their mass. Slow, laborious, and the odds are long.

The Stanford team built an algorithm called Peptide Predictor instead. It scanned the roughly 20,000 human protein-coding genes, looking for sequences where a specific enzyme — prohormone convertase 1/3, previously linked to human obesity — was likely to cut. The team filtered for proteins released outside cells, a hallmark of hormones, and required at least four cutting sites.

That left 373 prohormones. The algorithm predicted the enzyme could generate 2,683 distinct peptides from them.

"The algorithm was absolutely key to our findings," Svensson said.

A 12-amino-acid surprise

The researchers tested 100 peptides, including GLP-1, on lab-grown nerve cells. GLP-1 tripled neuronal activity compared with untreated controls. BRP — just 12 amino acids long, cut from a protein called BRINP2 — increased it tenfold.

Injected into muscle before feeding, it cut food intake by up to 50% over the following hour in mice and minipigs, which mirror human metabolism and eating patterns more closely than mice.

Notably, nearly all the weight the treated obese mice lost came from fat, not muscle. That matters, because muscle supports strength, mobility, and metabolic health — and muscle loss is a documented complaint with current weight-loss drugs. Nausea, another common GLP-1 drug side effect, can't be ruled in or out from animal work.

What does this change for patients?

Right now, nothing. BRP has never been tested in a person. The researchers still don't know which cell-surface receptors it binds, how long its effects last, or what dose might work. Peptides also break down quickly in the body, so a longer-lasting version may be needed to avoid frequent injections.

Svensson has cofounded Merrifield Therapeutics to push BRP toward clinical trials, and she and Coassolo hold patents on BRP peptides for metabolic disorders — a financial interest worth flagging. Collaborators included scientists from UC Berkeley, the University of Minnesota, and the University of British Columbia, with funding from the National Institutes of Health and several Stanford programs, among others.

"Nothing we've tested before has compared to semaglutide's ability to decrease appetite and body weight," Svensson said. "We are very eager to learn if it is safe and effective in humans."