Scientists at Stanford Medicine used artificial intelligence to comb through the human genome and surface a natural molecule that curbs appetite — one they say worked in animals without the nausea and other side effects common to Ozempic-style drugs.

How AI found it

The team built an algorithm called Peptide Predictor that scanned all roughly 20,000 human protein-coding genes for the signatures of hormone-like peptides, narrowing hundreds of thousands of possibilities to a shortlist and spotlighting a 12-amino-acid peptide named BRP, derived from a prohormone called BRINP2.

“The algorithm was absolutely key to our findings,” said senior author Katrin Svensson.

Why peptide hormones hide in plain sight

The search is possible because many hormones are not encoded as genes in their own right.

They are cut from larger precursor proteins — prohormones — by enzymes that recognise specific cleavage sites. A single precursor can yield several active fragments, and those fragments are short, sometimes only a handful of amino acids.

That makes them easy to overlook. Genome annotation focuses on protein-coding sequences, and a short peptide released from a larger protein is not annotated separately. Insulin, glucagon and many other hormones are produced this way, so the pattern is established — and the reasonable assumption has long been that more remain unidentified.

What the algorithm does is search for cleavage-site signatures and other features distinguishing genuine hormone precursors from ordinary proteins — a pattern-recognition problem across a defined dataset, which is exactly what such methods do well.

Why it might be gentler

The proposed advantage comes down to where the molecule acts.

Semaglutide hits receptors across the brain, gut and pancreas — which is part of why it can cause nausea and constipation. BRP appears to act more narrowly, in the hypothalamus, the brain’s appetite-and-metabolism control centre.

The reasoning is sound in principle. GLP-1 receptors are widely distributed because the natural hormone has several jobs — slowing gastric emptying, stimulating insulin release, signalling satiety — and a drug activating all of them produces effects beyond the one intended. Slowed gastric emptying is directly responsible for much of the nausea.

A peptide acting only where appetite is regulated would avoid those. Whether BRP is genuinely that selective in humans is the claim requiring testing.

What the animals showed

In lean mice and minipigs, it suppressed appetite and reduced food intake by up to 50%, with fat loss and without the muscle loss or nausea seen with GLP-1 drugs.

Minipigs are a meaningful addition. Rodent metabolism differs substantially from human, while pig physiology, gastrointestinal anatomy and body size are considerably closer — so a result reproducing across both species is more informative than a mouse result alone.

Testing in lean animals is a notable limitation, though. Appetite regulation differs in the obese state, and demonstrating reduced intake in a normal-weight animal is not the same as demonstrating weight loss in an obese one.

Assessing nausea in animals

The absence of nausea deserves scepticism on methodological grounds. Rodents cannot vomit, and nausea is a subjective experience an animal cannot report.

Researchers infer it from behaviour — conditioned taste aversion, or characteristic postures — which are indirect proxies. Pigs can vomit, which makes them the better model here, and even so absence of observable illness is weaker evidence than a person saying they feel fine.

The important caveat

The results, published in Nature, come from animals — human trials are only planned.

Why a natural peptide is not automatically safer

The framing of BRP as a natural molecule invites an inference worth examining, because it does not follow.

A peptide the body already produces is being given at doses far above what the body makes, at times the body would not make it, and in a distribution that bypasses normal regulation. Insulin is natural and lethal in excess; thyroid hormone is natural and dangerous when dosed wrongly.

Natural origin does say something useful about tolerability in one respect: the body possesses machinery to degrade and clear the molecule, and is unlikely to mount an immune response against something it produces itself. Both are genuine advantages over a foreign compound.

What it does not establish is that supraphysiological exposure is benign. If BRP has functions beyond appetite — and most peptide hormones do — sustained high levels would engage those too.

The GLP-1 drugs are the immediate precedent. GLP-1 is also a natural human peptide, and the drugs based on it produce substantial side effects precisely because they exceed physiological signalling.

Many promising weight-loss compounds look excellent in mice and disappoint in people, and no side effects in animals is not the same as safe in humans. Still, it is a striking demonstration of AI accelerating discovery. General information, not medical advice.