THE SCIENCE BEHIND
LEMBAS
Engineering bioactive peptides that pull their weight.
EXECUTIVE SUMMARY
Lembas engineers food-grade bioactive peptides that activate the body's native satiety biology - the same gut receptor network targeted by blockbuster GLP-1 drugs - but inside ingredients consumers can eat every day. Our proprietary discovery platform searches inside edible proteomes only, designs against a structurally mapped library of gut receptors, and validates every hit in the wet lab. Lead asset GLP-1 Edge™ (P197) has matched a pharmaceutical benchmark in preclinical efficacy, with no observed toxicity.
BY THE NUMBERS
10B+
Edible peptide sequences
100+
Gut GPCRs mapped
85%
Validation hit-rate
<12 MO
Discovery to clinic
-9.4
Kcal/mol binding (P197)
94.2%
Structural confidence (P197)
WHAT WE'VE BUILT
Edible Peptide Database: 10B+ food-derived sequences, pre-tagged for safety, allergenicity, solubility, gastric stability
Receptor Mapping Database: 100+ gut GPCRs structurally resolved; multi-axis design (GLP-1 + PYY + GIP)
Lembas AI Engine: generative design + active learning; closed-loop with the wet lab
GLP-1 Edge™ (P197): 10–20 aa food-derived peptide, ingredient-ready at 50–100 mg/serving
WHY IT MATTERS
GRAS-aligned by design: every candidate lives inside an edible protein
Activates the body's native satiety hormones in parallel, not just one receptor
No reformulation: Drop-in for bars, shakes, gummies, beverages
THE LEMBAS AI DISCOVERY ENGINE
At the core of the company is the Lembas AI Discovery Engine, an end-to-end computational and physical platform designed to systematically map, generate, and validate natural peptides. This platform operates on a rigid 5-stage closed loop, ensuring that computational predictions are rapidly grounded in biological reality, and that biological reality continuously makes the computation smarter.
SIDEBAR
WHAT IS A GCPR, IN PLAIN ENGLISH?
G Protein-Coupled Receptors (GPCRs) are the body's molecular 'antennas.' They sit on the outer surface of your cells, waiting to detect specific outside signals, which they then translate into actions inside the cell. Your gut is heavily lined with them. When a specific nutrient or peptide binds to the correct GPCR antenna, the gut cell releases powerful satiety hormones like GLP-1, PYY, and GIP into the bloodstream, telling the brain you are full. Lembas designs specific, natural peptides explicitly shaped to ring those exact molecular bells.
EDIBLE PEPTIDE DATABASE
Every Lembas program begins inside food. Stage 1 comprises our proprietary library of over 10 billion sequences mapped exhaustively across the proteomes of plants, livestock, dairy, and coffee. We do not merely catalog sequences; every entry is extensively computationally tagged. We annotate for GRAS alignment, predicted allergenicity risk, solubility profiles, and projected stability within gastric environments. This library is deeply queryable by structural motif, target receptor affinity, or source organism, ensuring the engine draws only from safe, biologically relevant starting materials.
LEMBAS AI ENGINE
Once a target is defined, the proprietary Lembas AI Engine is deployed. This is an advanced generative design system strictly trained on validated peptide–GPCR interactions. Utilizing iterative active learning, the AI engine proposes novel candidate sequences derived from our edible database, scores their binding potential against specifically targeted receptor pockets, and continuously refines its own architectural model with every data point returned from the wet lab. This iterative self-correction ensures the generative model constantly evolves its understanding of true biological binding dynamics.
RECEPTOR MAPPING DATABASE
Peptides cannot be optimized in a vacuum; they must be fitted to specific biological locks. Our Receptor Mapping Database contains high-resolution structural resolutions and binding-pocket annotations for over 100 specific gut GPCRs. Crucially, this encompasses the receptors controlling the GLP-1, PYY, GIP, leptin, and insulin physiological axes- the comprehensive network of native human satiety. By mapping the full array of receptors, the engine is empowered to design for multi-target activation rather than a pharmacological single-receptor pull.
IN-SILICO FILTERING
Before any capital or time is expended in physical testing, the AI Engine subjects every candidate sequence to five rigorous computational filtering gates simultaneously. Candidates are aggressively scored against toxicity risks, allergenicity homology, aqueous solubility, gastric stability (simulating survival through harsh gastric and intestinal fluid environments), and finally, manufacturability at commercial ingredient scale. Only sequences that clear all five gates proceed.
WET-LAB VALIDATION LOOP
The surviving fraction of computationally perfected candidates transitions to the physical lab. They are first tested in high-throughput in vitro STC-1 enteroendocrine cell assays to directly measure real-world hormone secretion. The top performers advance to in vivo DIO (Diet-Induced Obesity) rodent studies to confirm systemic metabolic efficacy. Crucially, this is a closed loop: every success, marginal result, and failure from the wet lab is fed directly back into the Lembas AI Engine.
This active-learning loop is the engine's defining advantage. Because every wet-lab result re-trains the model, the platform gets tangibly sharper, faster, and more predictive with every consecutive program it runs.
CASE STUDY: LEMBAS EDGE™ P197
The lead asset emerging from the Lembas platform is program designator P197, commercially known as GLP-1 Edge™. It represents the culmination of our 'food-grade by design' philosophy: an edible-protein-derived sequence featuring a stable 10-to-20 amino-acid backbone. It is highly water-soluble, highly stable, and ingredient-ready at an intended human dosage of 50 to 100 milligrams per serving.
P197 exerts its effects by activating the gut's native incretin axis through natural nutrient-sensing GPCRs. Critically, its mechanism is multi-axis. P197 engages receptors responsible for the release of GLP-1, PYY, and GIP in parallel. By orchestrating this combined endocrine response, it mirrors how a biologically optimal meal physiologically signals deep, sustained satiety to the brain.
-9.4
Kcal/molecule
94%
Predicted-to-validated rate against target GPCR
94.2%
Structural confidence
100k
Variants scored across Lembas Edge™ design campaign
Structural Binding: The exact amino acid sequence of P197 is withheld for IP reasons; the structural mechanism is published here for scientific transparency.
In silico modeling confirms that P197 adopts an exceptionally stable predicted binding mode deep within the 7-transmembrane pocket of the target gut GPCR. The molecule's architecture aligns perfectly with the receptor's orthosteric site, utilizing key contact residues to stabilize the active conformation of the receptor, thereby triggering the downstream intracellular signaling cascade that leads to hormone secretion.