Peptide classes: the 6 main families, and what backs each one

Why the family matters more than the name

A peptide’s family tells you more than its name. Family means a shared mechanism, and a shared mechanism means shared effects, shared side effects, and a shared evidence base. Once you can place a compound in a family, you can predict roughly how much human research exists before you read a single study.

Search for a list of peptides and you get an alphabet. BPC-157 sits next to bremelanotide, epitalon next to exenatide, one line of description each, every row the same width. That format is the problem, because it presents a compound with tens of thousands of trial participants behind it and a compound with none as the same kind of thing.

A peptide is a short chain of amino acids, usually weighing between 500 and 5,000 daltons (a dalton is the unit chemists use for the mass of a single molecule) [2]. That one definition stretches across everything. It covers insulin, 51 amino acids long, first isolated by Frederick Banting in 1921 and sold commercially from 1923 [2]. It also covers a three-amino-acid fragment being shipped from an unmarked warehouse today. More than 80 peptide drugs have reached the market since insulin, spanning diabetes, cancer, osteoporosis, multiple sclerosis, HIV infection and chronic pain [1]. Thirty-three non-insulin peptide drugs have been approved worldwide since 2000, and more than 170 peptides are in active clinical development [2].

What organises that sprawl is not the alphabet. It is the receptor, the specific lock on the surface of a cell that a peptide is shaped to fit. Peptides built for the same lock behave alike: similar benefits, similar side effects, and, because a receptor is what a drug company patents and funds trials around, a similar research history too. The metabolic family has trials enrolling thousands of people because one receptor attracted that investment. The repair family has almost nothing, because nobody funded it.

So the six families below are not a filing cabinet. Each one carries a rough answer to the only question a beginner actually needs: how much of this has been tested in people?

Metabolic peptides: the incretin family

Incretin peptides copy the gut hormones your body releases after a meal, which lowers blood sugar and blunts appetite. This is the best-evidenced family by a wide margin: semaglutide, tirzepatide and their relatives have been tested in trials enrolling thousands of people each, with published weight, blood sugar and cardiovascular results.

This family copies incretins, the hormones your gut releases when food arrives, which tell the pancreas to make insulin and tell the brain that the meal is over. Semaglutide, liraglutide, tirzepatide and retatrutide all work this way, and they are the reason the word peptide entered ordinary conversation.

The evidence here is not close to anything else on this page. In the STEP 1 trial, 1,961 adults with obesity and no diabetes were randomly assigned, two to one, to 68 weeks of once-weekly semaglutide or placebo alongside lifestyle support [3]. Mean body weight changed by -14.9% with semaglutide against -2.4% with placebo [3]. SURMOUNT-1 ran the same design for tirzepatide in 2,539 adults over 72 weeks, reporting -15.0%, -19.5% and -20.9% across the three doses against -3.1% for placebo [4].

Weight is the headline, but the trial that separates this family from every other one measured something harder. SELECT enrolled 17,604 patients aged 45 or older who already had cardiovascular disease and overweight or obesity but not diabetes, followed them for a mean of 39.8 months, and counted deaths from cardiovascular causes, heart attacks and strokes [5]. Those events occurred in 6.5% of the semaglutide group and 8.0% of the placebo group, a hazard ratio of 0.80 [5].

Keep that number in view for the rest of this article. It is what a mature evidence base looks like: tens of thousands of people, a hard outcome that nobody can talk themselves into feeling, and a result published where anyone can check it. No other family below has been tested at anything like that scale.

Growth hormone peptides

Growth hormone peptides do not contain growth hormone. They prod the pituitary gland into releasing more of your own, either by copying the releasing hormone above it or by copying ghrelin, the hunger signal. One member is approved for a narrow indication. The rest have small trials showing hormone levels move.

The second family is the one most often misdescribed. Sermorelin, CJC-1295, tesamorelin, ipamorelin and MK-677 contain no growth hormone at all. They act one step upstream, on the pituitary gland (the hormone control centre at the base of the brain), either by copying growth-hormone-releasing hormone or by copying ghrelin, the signal your stomach sends when it is empty. MK-677 is the odd one out, and a useful illustration: it is not a peptide at all but a small molecule, and it sits in this section anyway because it acts on the same receptor as ipamorelin. Grouping by receptor is what puts it here. Grouping by the alphabet never would.

Tesamorelin is the one with a real approval, and it is worth seeing what earning that took. 412 patients with HIV and abdominal fat accumulation were randomly assigned to a daily injection or placebo for 26 weeks; visceral fat measured on computed tomography fell by 15.2% on tesamorelin and rose by 5.0% on placebo [6]. A specific population, a specific problem, an imaging endpoint.

The rest of the family stops earlier. Two ascending-dose trials of CJC-1295 in healthy adults aged 21 to 61 showed growth hormone rising two- to ten-fold for six days or more after a single injection, with a half-life of 5.8 to 8.1 days [7]. That is a pharmacology result: the hormone moves. It is not a result about muscle, injury or aging. The closest anyone has come is a two-year trial of oral MK-677 in 65 healthy adults aged 60 to 81, where fat-free mass rose 1.1 kg against a 0.5 kg fall on placebo, and the authors said plainly that the study was underpowered to evaluate functional endpoints [8]. More lean mass on a scan, with no demonstrated change in what anyone could do.

Repair peptides

Repair peptides such as BPC-157 and TB-500 are the family with the widest gap between reputation and evidence. A 2026 sports medicine review puts it plainly: favorable tissue repair results in animal models, scarce rigorous human safety data, and potential for serious harm. No member of this family is approved anywhere.

Ask a gym or a forum which peptide to take for a bad tendon and you will be pointed at this family: BPC-157, TB-500 and thymosin beta-4. It is also the family with the thinnest floor under it.

A 2026 narrative review in Sports Medicine surveyed the peptides marketed directly to patients for musculoskeletal injury and performance, naming AOD-9604, BPC-157, CJC-1295, follistatin-344, GHK-Cu, ipamorelin, MOTS-c, sermorelin, SS-31, tesamorelin, thymosin beta-4 and TB-500. Its summary of the unapproved ones is the sentence to remember: favorable tissue repair and metabolic outcomes in animal models, but rigorous human safety data are scarce, and there is potential for serious harm [9]. The same review discusses the placebo effect as a mediator of these peptides’ apparent efficacy, and how social media amplifies it [9].

It is worth looking at what the human literature actually consists of, because the citation counts make it sound larger than it is. The most-cited human report on BPC-157 for joint pain is a retrospective chart review at a single clinic in Orlando, covering one year of records. Seventeen patients; sixteen reachable by telephone; twelve who received BPC-157 alone, of whom eleven reported meaningful improvement [10]. The authors state that no specific tools were used to measure function, quality of life, stiffness or daily activities [10].

That is not a criticism of those authors, who describe their own limits accurately and call for proper studies. It is a description of the family. Sixteen remembered phone calls is the human evidence base, against 17,604 randomly assigned patients one section above.

Skin peptides

Skin peptides are applied topically rather than injected, which changes the question entirely: the peptide has to cross the outer layer of skin before it can do anything. Two are worth knowing. Argireline interferes with the signal that contracts a muscle, and GHK-Cu is a copper-carrying fragment that falls with age.

Cosmetic peptides are a different argument from every other family here, because the obstacle is not the receptor but the delivery. A molecule rubbed on the face has to get through the stratum corneum (the dead, waterproof outer layer of skin) before any mechanism becomes relevant.

Argireline, also sold as acetyl hexapeptide-8, is patterned on a protein involved in releasing the signal that contracts a muscle. Its best-known controlled test randomly assigned 60 subjects three to one to argireline or placebo, applied to crow’s feet twice daily for four weeks. On the subjective assessment, total anti-wrinkle efficacy was 48.9% in the argireline group and 0% in the placebo group [11]. Sixty people for four weeks is a real trial and a small one, and the headline number is a rater’s judgement rather than a measurement.

GHK-Cu is a three-amino-acid fragment that carries copper, and the fact underneath its popularity is genuinely interesting: plasma levels are around 200 nanograms per millilitre at age 20 and about 80 by age 60 [12]. A cream containing it, applied to the faces of 71 women with photoaging for 12 weeks, increased skin density and thickness and reduced laxity [12]. Read that source knowing who wrote it: the lead author discovered GHK and works in skin-care research and development, which does not make the data wrong but does mean it wants independent replication [12].

Brain peptides

Brain peptides are the family where a large evidence base exists and still says no. Cerebrolysin, a porcine brain extract used for stroke across Russia, eastern Europe and China, has been through seven randomized trials and a Cochrane review. The review found no benefit on death and a possible increase in serious adverse events.

Semax, Selank, dihexa and DSIP get the attention online, but the informative member of this family is the one that has actually been tested at scale. Cerebrolysin is a mixture of low-molecular-weight peptides and amino acids derived from pig brain, and it is widely used for acute ischaemic stroke in Russia, eastern Europe, China and other post-Soviet countries [13].

A Cochrane review (Cochrane reviews pool every eligible trial and grade how much confidence the combined result deserves) updated in 2023 found seven randomized trials covering 1,773 participants [13]. On all-cause death, the pooled risk ratio was 0.96, with a confidence interval running from 0.65 to 1.41 across six trials and 1,689 participants, rated moderate certainty [13]. An interval that spans 1 like that means the data cannot distinguish a benefit from a harm. Meanwhile the total number of people with non-fatal serious adverse events was higher on cerebrolysin, risk ratio 2.39 [13].

Two further details make this the most useful case study on the page. None of the included studies reported poor functional outcome, quality of life, or time to restoration of capacity for work [13], meaning the outcomes patients care about were not measured. And the manufacturer supported three of the multicentre studies [13]. This is what it looks like when a peptide is genuinely popular, genuinely studied, and still does not clear the bar. Volume of research is not the same as weight of evidence.

Longevity peptides

Longevity peptides split into two halves with very different foundations. Mitochondrial peptides such as MOTS-c were discovered by reading the mitochondrial genome and have real mechanistic work behind them, almost all in mice. Khavinson bioregulators such as epitalon and thymalin come from one research group, reviewing its own long-term studies.

The last family is really two families that get sold as one, and telling them apart is the most useful thing a beginner can learn here.

The first half is mitochondrial peptides. Mitochondria, the compartments inside your cells that produce energy, carry their own small loop of DNA, and researchers found short stretches of it that encode peptides [15]. MOTS-c is one: a 16-amino-acid peptide encoded within the mitochondrial 12S rRNA, whose main target organ appears to be skeletal muscle [15]. Treating mice with it prevented both age-dependent and high-fat-diet-induced insulin resistance, and prevented diet-induced obesity [15]. That is a real and well-made discovery, and every outcome in that sentence happened in mice.

The second half is the Khavinson bioregulators: epitalon, thymalin, vilon, cortexin and their relatives, very short peptides developed by one group in St Petersburg. The clinical case for them rests largely on a 2013 review in a gerontology journal that, by its own description, summarises the results of long-term studies by the review’s own authors on Timalin, Thymogen, Vilon, Epithalamin, Prostatilen, Cortexin and Retinalamin [14].

A group reviewing its own work is not fraud and it is not nothing. It is simply the weakest arrangement evidence can take, because the check that makes a finding trustworthy, someone unconnected trying it and reporting what happened, has not been run. Compare it to the cerebrolysin entry above, where independent reviewers pooled everything and reached an unwelcome answer. That contrast is the whole point of grouping peptides by family.

Placing a name you have never seen

Three questions place almost any peptide. What receptor does it act on, which gives you the family. Is any member of that family approved by a regulator, which tells you whether the mechanism has survived scrutiny. And does the human evidence consist of randomized trials, a chart review, or mice?

New names appear constantly, and the point of a map is that you do not need a new entry for each one. Three questions place almost anything.

First, what receptor does it act on? That lands you in a family, and the family carries its own expectations about side effects. Anything incretin-like will involve nausea and slowed stomach emptying. Anything working through the pituitary will raise insulin-like growth factor 1, which is why that family is prohibited in competitive sport at all times.

Second, is any member of the family approved by a regulator for anything? Semaglutide and tesamorelin are approved, so their mechanisms have been through a process that looked for harms. Nothing in the repair family has [9], which means no regulator has ever weighed its risks against its benefits for anyone.

Third, and most important, what shape is the human evidence? Not how many studies, but what kind. Randomized and placebo-controlled, with an outcome that does not depend on anyone’s opinion, is one thing. A retrospective chart review of sixteen remembered phone calls is another [10]. A review by the authors of the studies it reviews is a third [14]. Mice are a fourth [15].

Six families is a map, not a census. Melanocortins such as bremelanotide and the melanotans sit outside it, as do antimicrobial peptides like LL-37 and the hospital families used in oncology. The peptide field as a whole spans urology, respiratory medicine, pain, oncology, metabolic disease, cardiovascular disease and antimicrobial use [2]. But the six above cover almost everything sold to consumers, and the habit is what transfers: find the family first, then ask what it is standing on.