Stanford AI Screen Identifies Hypothalamus-Selective Peptide BRP as GLP-1 Alternative

The team trained a neural network on mass-spectrometry datasets from human and mouse tissues to predict which of thousands of peptide fragments would bind neural receptors involved in energy balance. Top candidates were synthesized and injected into diet-induced obese mice; BRP produced 15-20 percent weight loss over 14 days with no detectable slowing of gastric emptying or change in blood glucose, unlike semaglutide controls. Single-nucleus RNA-seq showed BRP activated a distinct subset of POMC and AgRP neurons confined to the arcuate nucleus, explaining the narrower side-effect profile observed.

Current GLP-1 agonists achieve broad efficacy because their receptors are expressed in gut, pancreas and brainstem, driving nausea in up to 44 percent of patients and measurable lean-mass loss. BRP's anatomic restriction could therefore support longer-term adherence and reduce sarcopenia risk, especially in older adults. The work also illustrates how AI can shortcut traditional peptidomics by prioritizing functional cleavage sites rather than abundance alone.

Svensson has launched a spin-out to manufacture GMP-grade BRP and file an IND. Planned first-in-human studies will test ascending doses in healthy volunteers followed by a 12-week proof-of-concept trial in adults with BMI 30-40; endpoints include safety, 24-hour caloric intake, and fMRI hypothalamic activation. Success would validate AI-guided peptide mining as a general platform for tissue-selective endocrine drugs.