What is a peptide? A biology and chemistry primer

A peptide is a short string of two to fifty amino acids joined by covalent bonds called peptide bonds ^[1]. That's it. Anything longer and we usually call it a protein. Anything not connected by peptide bonds isn't a peptide, even if it shares some letters with one on a marketing page.

Why this matters in practice: the cellular machinery that recognizes peptides - receptors on cell surfaces, enzymes that chew them up, transporters that move them around - was built by evolution to respond to these specific molecular shapes. When a drug company makes a peptide therapeutic, they're not inventing a new key. They're filing down a copy of one your body already uses.

Twenty amino acids do the work. Almost every peptide and protein in human biology is built from the same short list. Each one shares the same skeleton - a central carbon, an amino group on one side, a carboxyl group on the other, a hydrogen up top - and then a side chain that varies from one to the next.

The side chain is the whole story. That's what makes glycine different from tryptophan, and tryptophan different from cysteine.

Think of the side chain as the personality. Some carry charge. Others are hydrophobic and want to hide from water, tucking themselves into the core of the folded protein where the solvent can't reach them. A few carry sulfur atoms that can throw bridges across to other parts of the chain.

The order matters.

String those twenty letters together in one sequence and you get a hormone that binds a receptor for fifteen minutes. Shuffle them, and in practice you get a misfolded blob that gets chewed up by a protease before it leaves the cell. Shape, stability, what receptor the molecule fits into, how long it lingers in your bloodstream - all of it cascades from sequence.

Twenty letters. That's a smaller alphabet than English. The combinatorial space is still effectively infinite.

Two amino acids meet. The carboxyl group of one reacts with the amino group of the next. A water molecule falls out. What remains is a covalent bond - the peptide bond - connecting the carbon of the first amino acid to the nitrogen of the second.

Biochemists call this a condensation reaction, or dehydration synthesis, because the system loses one water per bond formed. In a living cell, this process runs on ATP and a piece of molecular machinery called the ribosome, which reads messenger RNA and assembles peptide chains amino-terminus first. The reverse reaction - adding water back to break the bond - is called hydrolysis, and it's what your digestive enzymes do to dietary protein.

A small but important quirk: the peptide bond isn't a simple single bond. Its electrons are partly shared with the neighboring carbonyl oxygen, giving it partial double-bond character. That restricts rotation. Which means the backbone of every peptide is locally flat and rigid in predictable ways, and that rigidity is the geometric reason α-helices and β-sheets exist at all.

Once an amino acid is inside the chain, biochemists stop calling it an amino acid. They call it a residue, because something - a water molecule - was left behind in the process. Counting residues is how peptide chemistry keeps score.

So where does a peptide end and a protein begin? The textbook answer is around 50 amino acid residues. The honest answer is that there's no hard line, and different fields draw it differently.

A dipeptide has two residues. An oligopeptide runs roughly two to twenty. Polypeptide usually means more than twenty, though some references push that cutoff to fifty. Protein is conventionally reserved for chains that fold into stable three-dimensional structures, which generally requires at least forty or fifty residues of material to work with. Insulin sits right on the border at 51 residues across two chains, weighing about 5,800 daltons, and gets classified as either the smallest protein or one of the largest peptides depending on who is writing ^[2].

Why does the line wobble? Because the definition was never really about length. It was about behavior. Short chains tend not to fold into stable, repeatable 3D structures on their own. Longer chains do. The 50-residue rule is a useful heuristic for that transition, not a law of nature. For regulatory and manufacturing purposes the distinction matters quite a bit - smaller peptides often get reviewed by the FDA as drugs under an NDA, while larger ones cross into biologics territory under a BLA, with different rules around manufacturing and characterization.

A peptide's sequence is one-dimensional. Its biological activity is three-dimensional. The same residues in the same order can fold differently under different conditions and behave like a different molecule entirely.

Receptors don't read sequences. They feel shapes. A GLP-1 receptor on a pancreatic beta cell will accept semaglutide because semaglutide's folded conformation presents the right contact points in the right places - not because the receptor inspects the sequence and approves it. This is why a single amino acid substitution can sometimes destroy activity completely, and why other substitutions, in positions that don't touch the receptor, are tolerated without much consequence.

It's also why peptides are harder to manufacture than the chemistry suggests. The molecule has to be made cleanly, folded correctly, and stay folded long enough to do its job. A peptide that arrives at the receptor in the wrong conformation is, functionally, a different drug.

The therapeutic peptide universe is broader than the GLP-1 conversation suggests, and worth understanding even if you only ever care about one corner of it.

Metabolic and incretin peptides. GLP-1 is the family everyone's heard of now: semaglutide, tirzepatide, the older exenatide, and a pipeline of dual and triple agonists that hit GIP and glucagon receptors too. They mimic gut hormones ^[3]. Specifically, the ones your intestinal L-cells already squirt out after a meal, which is why the mechanism feels less like a foreign drug than like turning up a dial the body was already running.

Hormones. Insulin. Oxytocin, vasopressin, parathyroid hormone, calcitonin - the oldest category in the peptide world, and still one of the biggest by prescription volume, because insulin alone props up most of that number. In practice these are direct copies, or near-copies, of hormones your own body makes.

Neuropeptides. Substance P, CGRP, the melanocortins (relevant to PT-141 / bremelanotide for sexual function), ghrelin. These act on the nervous system and on receptors expressed throughout the body. Some, like CGRP-targeting drugs, anchor major franchises in migraine medicine.

Growth factors and growth-hormone-related peptides. Sermorelin, ipamorelin, CJC-1295, tesamorelin. These prod the pituitary to release growth hormone, or imitate fragments of growth hormone itself. Tesamorelin is the only one in the group with full FDA approval for an indication (visceral fat in HIV-associated lipodystrophy).

Antimicrobial peptides. A diverse and ancient class, present in everything from human skin to frog secretions. Some neuropeptides - PACAP, VIP, α-MSH - turn out to have direct antimicrobial activity to their signaling roles ^[4]. Therapeutic translation has been slow and uneven, but the category exists.

Tissue-repair peptides. BPC-157 is the famous example. The clinical evidence base in humans remains thin, and the regulatory status is messy, which we'll get to.

Once chemists could read a peptide's sequence, the next question was obvious: can we improve on it?

Two minutes. That's the half-life of natural GLP-1 once it hits human blood. An enzyme called DPP-4 snips it apart almost the moment it's released, which is elegant if you're a body trying to fine-tune a meal-timed satiety signal, but a disaster if you're trying to build a drug, because nobody wants an injection every fifteen minutes. So developers went hunting for tweaks.

Exenatide, the first GLP-1 receptor agonist approved (2005), was lifted from the saliva of a Gila monster - a peptide called exendin-4 that happens to resemble human GLP-1 closely enough to activate the same receptor while resisting DPP-4. Liraglutide added a fatty acid chain that lets the molecule hitchhike on albumin in the bloodstream, extending half-life to about thirteen hours. Semaglutide refined the same trick, plus a couple of amino acid substitutions, pushing half-life to roughly a week.

The standard toolkit for stabilizing a peptide is well established now ^[5]. Swap an L-amino acid for its mirror-image D-form at the cleavage site so enzymes don't recognize it. Cyclize the molecule - join the two ends - so there's no free terminus to attack. Attach a lipid or a polyethylene glycol chain to slow renal clearance. Make backbone substitutions that aren't quite peptide bonds anymore but behave similarly in 3D space.

Each modification trades something. Greater stability often comes at the cost of receptor affinity, or oral bioavailability, or manufacturing complexity. The art of peptide drug design is the trade-off management.

Frederick Banting and Charles Best isolated insulin from dog pancreases in 1921. Within a year, a fourteen-year-old boy named Leonard Thompson, dying of type 1 diabetes in a Toronto hospital, became the first human treated. He lived another thirteen years. That was the beginning of peptide medicine.

For the next sixty years, therapeutic peptides came from animals. Insulin from pigs and cattle. Other hormones extracted from cadaver pituitaries - a practice that ended after the cadaver-derived growth hormone supply was linked to Creutzfeldt-Jakob disease in recipients. The whole industry needed a different path.

Recombinant DNA technology, commercialized in the late 1970s and rolled out medically in 1982 with the first recombinant human insulin, gave it one. Now you could engineer bacteria or yeast to manufacture human peptide sequences at scale, without slaughterhouses or organ banks. The transition took roughly a decade. By the early 1990s, most therapeutic peptide hormones were recombinant.

The GLP-1 era started in 1987, when researchers characterized the insulin-stimulating effects of the hormone for the first time ^[6]. Exenatide arrived in 2005, liraglutide in 2010, semaglutide in 2017, and then the big shift: semaglutide for weight loss in 2021, and tirzepatide - a dual GLP-1/GIP agonist - for obesity in 2023. More than eighty peptide drugs have been approved globally by recent counts. The pipeline is denser than at any point in the field's history.

A small-molecule drug - aspirin, metformin, most pills in a pharmacy - typically weighs under 500 daltons and is built by traditional medicinal chemistry. A peptide drug usually weighs 500 to 5,000 daltons and is built by stitching amino acids together, either chemically or biologically.

The differences matter to patients in concrete ways.

Specificity. Peptides tend to bind their targets more selectively than small molecules. They're shaped to fit one receptor, and they engage a larger contact surface. Off-target effects are often fewer, though never zero.

Immunogenicity. Larger molecules carry more risk that the immune system will recognize them as foreign and mount a response. Modern human-sequence peptides have largely solved this for common drugs. Older animal-derived insulin caused real problems for some patients; current recombinant human insulin rarely does.

Delivery. Small molecules can usually be swallowed. Most peptides can't. Injection, nasal spray, or specialized oral formulations dominate.

Manufacturing. Small molecules are made by chemical synthesis at scale, often cheaply. Peptides require either solid-phase synthesis (a precise but expensive multi-step chemical process) or recombinant biology (cell lines, fermentation tanks, purification trains). Cost per dose runs higher. Quality control runs harder. Even a single amino acid out of place can change activity or trigger an immune response.

Regulation. Smaller synthetic peptides often go through the same FDA pathway as small molecules. Larger ones, and most recombinant ones, fall under biologics regulation, which is a different and more demanding review process.

A grown-up read of the field admits what the science still doesn't know.

Long-term effects of chronic GLP-1 agonism in people without diabetes are still being mapped. The drugs work. Cardiovascular outcomes data is genuinely strong, but the very long horizon - what twenty or thirty years of weekly receptor saturation does to a body - is, by definition, not yet in the dataset.

Short answer: people taking these drugs now are the longest study.

The compounded peptide market is its own story. BPC-157, CJC-1295, GHRP-2, and similar peptides are widely sold and injected without FDA approval for any indication. The human clinical evidence ranges from preliminary to nonexistent. The FDA has moved to restrict compounding of several of these on safety grounds. Use cases exist; the published clinical evidence runs well behind the marketing.

Oral peptide delivery remains hard. Oral semaglutide works, sort of, with absorption enhancers and a particular dosing protocol. Generalizing that to other peptides has not been easy. Most therapeutic peptides will remain injectable for the foreseeable future.

And the residue-count rule, the one we opened with, is still a convention. Insulin is a protein when a regulator needs it to be and a peptide when a marketer does. The molecule doesn't care which side of fifty residues someone decides to draw the line. The biology is the same either way.

Knowing that's the difference between reading a peptide therapy page critically and not.