Semaglutide's structure: Aib, fatty acid and half-life

Semaglutide's structure is the 31-amino-acid human GLP-1(7-37) peptide with three deliberate changes: the alanine at position 8 is swapped for Aib (alpha-aminoisobutyric acid) to resist the enzyme DPP-4, the lysine at position 34 is swapped for arginine, and the lysine at position 26 carries a side chain made of a linker plus a C18 fatty diacid that binds albumin. Together those changes turn a short-lived hormone into a molecule with an elimination half-life of about one week, according to the FDA-approved Ozempic label. This page walks through each piece and explains why it matters.

The starting point: native GLP-1

Glucagon-like peptide-1 is a gut hormone. In its active form, GLP-1(7-37), it is a short peptide that the enzyme dipeptidyl peptidase-4 (DPP-4) cuts at its N-terminus. The Novo Nordisk team that designed semaglutide described native GLP-1 as "rapidly metabolized by DPP-4," which is why it is useless as a once-weekly molecule on its own. The design job was to keep receptor activity while fixing two problems: enzyme cleavage and fast clearance.

Semaglutide was built on lessons from liraglutide, Novo Nordisk's earlier daily GLP-1 analogue. Both molecules attach a fatty acid to lysine 26, but semaglutide changes the backbone, the linker and the fatty acid itself.

Semaglutide's chemical structure, piece by piece

Aib at position 8: the DPP-4 shield

Aib is a non-standard amino acid with two methyl groups on its alpha carbon. Placing it at position 8, the second residue of the active peptide, makes the N-terminus a poor substrate for DPP-4. The FDA label states that semaglutide "is modified in position 8 to provide stabilization against degradation" by DPP-4. In the 2015 discovery paper by Lau and colleagues in the Journal of Medicinal Chemistry, an otherwise similar analogue that kept alanine at position 8 had a much shorter half-life in rats, which the authors attributed to DPP-4 cleavage.

The peptide backbone with these two substitutions was crystallized bound to the extracellular domain of the GLP-1 receptor. The Protein Data Bank entry 4ZGM labels it "8Aib,34R-GLP-1(7-37)-OH" and resolves it at 1.80 Å.

Arg34: making the chemistry site-specific

Native GLP-1(7-37) has two lysines, at positions 26 and 34. If both were free, a fatty acid could attach at either one, or both, giving a mixture. Replacing lysine 34 with arginine, a similarly charged residue without a reactive side-chain amine, leaves lysine 26 as the only attachment point. The label calls this "a minor modification ... to ensure the attachment of only one fatty di-acid." Lau and colleagues reported that Arg34 had no apparent effect on receptor binding.

The semaglutide side chain: linker plus C18 fatty diacid

The semaglutide side chain is what most people mean when they ask about its structure. It hangs off lysine 26 and has three parts:

• Two OEG units. Short, hydrophilic ethylene-glycol-based amino acid spacers that put distance between the peptide and the lipid.
• A gamma-glutamic acid (γGlu). Carried over from liraglutide, where it supplied an acidic group.
• A C18 fatty diacid. An 18-carbon chain with a carboxylic acid at each end. Liraglutide uses a 16-carbon monoacid (palmitic acid).

In the discovery work, albumin affinity rose with diacid length, and C18 gave the best potency among the diacids tested. The linker choice also strongly affected albumin binding and potency. The result was a trade-off the team accepted: semaglutide binds the GLP-1 receptor about three times less tightly than liraglutide (0.38 nM vs 0.11 nM in their assay), but binds albumin far more strongly, about 5.6 times more by analytical ultracentrifugation.

How the structure produces a one-week half-life

The FDA label is direct: "The principal mechanism of protraction resulting in the long half-life of semaglutide is albumin binding." Semaglutide is more than 99% bound to plasma albumin. Binding to albumin, a large and long-lived plasma protein, leaves only a small free fraction of the peptide at any moment. The Aib substitution handles the main enzymatic weak point that remains.

Breakdown follows the structure too. The label describes metabolism by proteolytic cleavage of the peptide backbone followed by sequential beta-oxidation of the fatty acid side chain, with only about 3% of the dose appearing intact in urine.

Licensed medicines vs research-grade semaglutide

Semaglutide is the active ingredient in Novo Nordisk's Ozempic and Wegovy, which are approved medicines with defined manufacturing, sterility and quality controls. The US label notes the peptide backbone is produced by yeast fermentation. A vial of research-grade semaglutide powder is not Ozempic or Wegovy and is not approved for human use. Its identity, purity and counter-ion depend entirely on the supplier, which is why a certificate of analysis you can verify matters for any lab work. For the regulatory side, see our peptide legal guide and FDA warning letters tracker.

Using the structure at the bench

For in-vitro work, the molecular weight is what you need. At 4113.58 g/mol, 1 mg of semaglutide (free peptide basis) dissolved to 1 mL gives roughly 243 µM. Check whether your certificate reports net peptide content or a salt form, since the counter-ion changes the mass you weigh. Our peptide calculator handles the molarity math, and the semaglutide reference page collects identifiers in one place.

To see how semaglutide's single-receptor design compares with dual and triple agonists, read semaglutide vs tirzepatide vs retatrutide. Tirzepatide, the active ingredient in Eli Lilly's Mounjaro and Zepbound, is covered on our tirzepatide page.

Frequently asked questions

Sources

1. Ozempic (semaglutide) US prescribing information, sections 11 and 12 (FDA label text)
2. Ozempic Summary of Product Characteristics, section 5 (UK, electronic medicines compendium)
3. Lau J et al. Discovery of the once-weekly GLP-1 analogue semaglutide. J Med Chem 2015;58:7370–7380
4. RCSB Protein Data Bank: 4ZGM, semaglutide peptide backbone with GLP-1 receptor extracellular domain (2015)