Skip to main
Peptide Broker

RESEARCH PEPTIDE FUNDAMENTALS / THE MATRIX

Four Classes, Read Against Each Other

Where each class acts, how long each molecule lasts, what kind of study exists to support it, and which single caution defines it.

The short version

Put side by side, these four peptides share a chemical description and almost nothing else. Comparing them on effectiveness would be meaningless — they are not competing for the same job. What can usefully be compared is structure: where each one acts, how long it stays in the body, what sort of evidence exists behind it, and what its main documented risk is.

Read that way, four patterns emerge. Where the receptor lives predicts the side effects better than the marketing does. The size and modification of the molecule predicts how often it has to be given. The class's development history predicts what kind of study a reader will find. And regulatory standing varies so widely across these four that the shared label ‘peptide' actively misleads.

The matrix below is the compressed version of the four compound briefs. Every figure in it is drawn from a cited study on the relevant page.

The class matrix

IpamorelinGHK-CuPT-141Tirzepatide
ClassGrowth hormone secretagogueCopper-binding matrix tripeptideMelanocortin receptor agonistDual incretin mimetic
Acts onGHS-R1a on pituitary somatotrophsFibroblasts, keratinocytes, extracellular matrixMC4R/MC3R in hypothalamus and limbic systemGIP and GLP-1 receptors
Chain length5 residues3 residues plus a copper ion7 residues, cyclic39 residues plus a C20 fatty diacid
OriginSynthetic, derived from GHRP-1Endogenous, found in type I collagenSynthetic analogue of alpha-MSHSynthetic, based on native GIP
Studied mainly forGrowth hormone release; postoperative ileusSkin matrix synthesis; hair growthHypoactive sexual desire disorderType 2 diabetes; obesity
Strongest human dataOne Phase 2 RCT that missed its endpoint (n=114) [3]Small topical trials; one 45-man combination trial [8]Two Phase 3 RCTs (n=1,267) plus a 52-week extension [13][14]Multiple 72-week Phase 3 RCTs (n=2,539; n=751) [21][18]
Measured half-life~2 hours [4]Not established in humans by any systemic route~2.7 hours (range 1.9-4.0) [15]Long enough for once-weekly dosing in trials [21][22]
Regulatory standingNot approved anywhere; prohibited in sportLegal cosmetic ingredient topically; unapproved injectableApproved for HSDD in premenopausal women [15]Approved for type 2 diabetes [19] and later indications
Signature cautionLong-term human safety uncharacterisedSystemic use unstudied; complex integrity is a safety parameterNausea and transient blood-pressure rise [14][15]Gastrointestinal intolerance during escalation; biliary risk [20]

Where the receptor lives predicts the side effects

The most transferable idea in this briefing is that adverse effects are rarely surprises. They are the receptor map, read honestly.

Melanocortin receptors sit in appetite circuitry and on pigment-producing cells as well as in the circuits governing desire. So a compound targeting MC4R for desire arrives with nausea reported at 40.4% over long-term use, flushing at 20.6%, headache at 12.0% [14], and darkening of skin, gums and moles with frequent dosing. None of that is off-target in any meaningful sense; it is the same target, in other tissue.

GHS-R1a sits on the pituitary but also on enteric and vagal neurons and in hypothalamic appetite circuitry. So a secretagogue reported for sleep and recovery also comes with appetite stimulation as a class property, and inherits the whole risk profile of the growth hormone axis it amplifies — insulin resistance, fluid retention, the theoretical IGF-1 and proliferation concern.

Incretin receptors sit in the pancreas, gut and hindbrain. So the dominant adverse events are gastrointestinal, concentrated during dose escalation [21][22], with delayed gastric emptying producing knock-on effects on other oral medicines and on pre-procedure stomach contents.

GHK-Cu has no such receptor address at all, which is precisely why its adverse profile is different in kind: local irritation, pigment change, product-stability failures. A compound that acts where it is applied has problems that are local.

Molecular architecture predicts the dosing shape

Chain length and chemical modification predict how often a molecule must be given more reliably than potency does, and the four here form an unusually clean progression.

Three residues and a metal ion, applied to skin, with no established systemic pharmacokinetics in humans by any route — GHK-Cu is dosed as a topical, and its central research problem is not duration but whether it crosses the stratum corneum at all, with a calculated partition coefficient of -2.24 for the free peptide [6].

Five residues, several of them in D-configuration to resist proteases, clearing with a terminal half-life of approximately two hours and producing a growth hormone peak around forty minutes after dosing [4] — a pulse, by design.

Seven residues, cyclised through a lactam bridge, with a terminal half-life near 2.7 hours [15] — long enough for an as-needed product with a labelled cap on frequency.

Thirty-nine residues carrying a C20 fatty diacid engineered to bind albumin, which supports once-weekly administration across the Phase 3 programme [21][22]. Every step up in size and modification buys duration, and duration is what decides whether a compound is a pulse, an as-needed dose, or a weekly injection.

Evidence maturity is the widest gap of all

If a reader takes only one comparison away, it should be this one, because it is the axis on which these four differ most and the axis least visible from a product description.

At one end sits the incretin class, with a 72-week trial in 2,539 adults [21], a head-to-head against another approved peptide in 751 adults [18], a 40-week diabetes trial in 1,879 adults [22], and a dedicated safety meta-analysis pooling 9,871 participants across nine trials [20]. Questions here are answered with confidence intervals.

Next sits the melanocortin class: two identical Phase 3 trials in 1,267 women [13] with a 52-week extension in 684 [14], a mechanistic imaging study in 31 [12], and an approval. Substantial and real — but with an unresolved argument about whether the measured change in a questionnaire score is a change that matters.

Third sits the matrix tripeptide, whose profile is inverted: gene-expression work reporting modulation of about 31.2% of human genes at a fifty-percent threshold [7], quantified skin permeation [10], and a clinical record of small topical studies plus one 45-man trial of a combination product [8].

Last sits the secretagogue class, whose only published randomised human trial missed its primary endpoint [3] and whose most recent published in-vivo work is a ferret study [1]. This is not an ordering of value — a cosmetic ingredient does not need a 72-week outcome trial. It is an ordering of confidence, and confidence is what a reader is actually trying to calibrate.

What class does not tell a reader

Two limits on this whole exercise are worth stating plainly, because a taxonomy that oversells itself is its own failure mode.

First, class predicts the shape of the risk, not its magnitude in any individual. Knowing that melanocortin agonism produces nausea does not say who will be among the roughly forty percent affected [14], and knowing that incretin therapy causes gastrointestinal effects during escalation does not say how severe they will be in one person. The literature describes populations. It does not describe readers.

Second, class says nothing about what is actually in a vial. Ipamorelin in circulation is research-grade material of variable purity from unregulated suppliers. Material sold as ‘PT-141 research chemical' has no verification of identity, purity or concentration. Compounded tirzepatide proliferated during a shortage period and drew regulatory concern about quality and identity. And GHK-Cu's activity depends on the copper remaining properly bound, so a degraded product may simply not be the compound the studies tested. Pharmacology assumes the molecule is what the label says. That assumption is the weakest link in this entire field, and no amount of class knowledge repairs it.