What Is a Triple Agonist Peptide? GLP-1, GIP, and Glucagon Explained

Potency and efficacy sit on different axes

The most common error in popular write-ups about multi-receptor peptides is treating one number as though it summarized everything a receptor assay found. It does not, because a concentration-response experiment produces at least two independent quantities and they can move in opposite directions.

Plot the log of agonist concentration on the horizontal axis and the measured response on the vertical axis, and a well-behaved agonist traces a sigmoid. Potency describes where that sigmoid sits horizontally. It is usually reported as EC50, the concentration at which the response reaches half of its own maximum in that experiment. Efficacy describes how high the sigmoid climbs. It is usually reported as Emax, the plateau of the response, and it is almost always expressed as a percentage of the plateau reached by a reference agonist run on the same plate.

Those two quantities are not versions of the same idea. A compound can be left-shifted relative to another, meaning lower EC50 and higher potency, while reaching a visibly lower ceiling, meaning lower Emax and lower efficacy. The reverse combination is equally possible. Writing that describes one compound as stronger than another has usually collapsed both axes into a single word, and that word cannot be recovered back into either measurement.

The deeper point is what each quantity is a property of. An EC50 is not a property of the compound. It is a property of the compound in a specified system, and the system includes the receptor, the species variant of that receptor, the host cell line, how many copies of the receptor that line expresses, the signal being read, the detection chemistry, the incubation time, the buffer composition, and the reference agonist used to normalize the y-axis. Drop any of those from the description and the number stops being reproducible in principle, not merely in practice.

There is also a formatting habit worth recognizing. Many papers report pEC50, the negative base-ten logarithm of the molar EC50, because the fitting is done in log space and error is better behaved there. A pEC50 rises as potency rises, which is the opposite direction to an EC50. Secondary summaries that mix the two conventions in one table have produced comparisons that are backward, and the only defense is checking the units and the direction before reading the ranking.

One further quantity gets quoted as if it were a measurement when it is actually a derived figure. A potency ratio between two compounds, or between two receptors for one compound, carries the uncertainty of both underlying fits. Two EC50 values that each have a wide confidence interval produce a ratio with a wider one, and a ratio printed as a bare integer has hidden that entirely.

What each reported quantity measures and how it is commonly misread

A useful discipline is to refuse to read any single agonism figure without asking which axis it belongs to and what the other axis did. If a summary reports a potency value and no ceiling, half the experiment is missing. If it reports a ceiling with no reference agonist named, the percentage has no denominator and cannot be checked. Both omissions are extremely common in material written downstream of the original papers, and neither is repaired by adding more decimal places.

What a concentration-response curve actually contains

Behind every EC50 is a fitted curve, and behind every fitted curve is a set of choices about concentrations, replicates, and the model being fitted. The number inherits every one of them.

The standard approach is a four-parameter logistic fit. The four parameters are the lower plateau, the upper plateau, the midpoint in log concentration space, and the Hill slope that sets how steeply the curve rises between the two plateaus. Only one of those four is normally quoted downstream. The midpoint becomes the EC50 and the other three vanish, even though the midpoint is only interpretable when the two plateaus are actually defined by data.

That is where most weak curves fail. If the highest concentration tested has not yet flattened the response, the upper plateau is not measured but extrapolated, and the fitting routine will still return an EC50 with a tidy confidence interval because the algorithm has no way to know the plateau was guessed. Curves that stop climbing only because the concentration series ran out produce EC50 values that are systematically unreliable, and the tell is a top parameter with a wide confidence interval or a curve whose last two points are still rising. A published figure showing the actual points lets a reader check this; a table of EC50 values with no figure does not.

The Hill slope carries information that is routinely discarded. A slope near one is what a simple single-site interaction predicts. A markedly steeper or shallower slope suggests something else is going on, such as multiple binding sites, cooperativity, a mixed population of receptor states, or an artifact of the detection chemistry saturating. None of that is visible once the curve has been reduced to one midpoint.

Replicate structure matters as much as the fit. Wells on the same plate, sharing a cell passage, a reagent lot, and a plate reader run, are technical replicates. They tell you about pipetting consistency and almost nothing about reproducibility. Independent experiments on separate days with separate cell passages are biological replicates, and the spread across those is what an error bar should represent. A reported EC50 with an impressively tight interval that turns out to be technical replicates within one plate is precise and not accurate.

There is also an arithmetic trap in how replicate EC50 values are combined. EC50 values are approximately log-normally distributed, so the correct summary across experiments is the mean of the pEC50 values, converted back if needed, which yields a geometric mean. Taking an arithmetic mean of raw EC50 values lets a single high outlier drag the average upward and produces a summary that is biased against the compound. This is not a hypothetical error; it appears whenever numbers are lifted out of papers and re-averaged in a spreadsheet.

The fitted parameters, and how a weak curve corrupts each one

None of this makes fitted potency values useless. It makes them conditional. A curve with both plateaus defined by real points, a free Hill slope, and error bars drawn across independent runs supports a potency claim that a reader can evaluate. A bare number in a summary table supports the weaker claim that someone once fitted something. The distance between those two situations is invisible unless the underlying figure is available, which is a good reason to treat secondary compilations of receptor data as pointers to papers rather than as data in themselves.

Why an EC50 does not transfer between publications

Two competent groups can publish EC50 values for the same molecule at the same receptor that differ by an order of magnitude, and both can be correct. This is not a scandal and it is not sloppiness. It is a direct consequence of the fact that the number describes a system rather than a molecule, and the two groups built different systems.

Receptor expression level is the largest single lever. A recombinant line engineered to express a receptor at high copy number has spare receptors, which means a submaximal fraction of occupied receptors is enough to drive a maximal downstream response. The practical result is a left-shifted curve: the EC50 falls, sometimes dramatically, without any change in how tightly the ligand binds. The same construct in a line with modest expression shifts right. Two labs using different stable clones of the same nominal cell line will not match each other, and neither will match a line with endogenous receptor.

The host cell background contributes independently. The complement of G proteins, the level of receptor kinase activity, the phosphodiesterase activity that degrades cyclic AMP, and any endogenous receptors of the same family all shape what the readout sees. Assays that include a phosphodiesterase inhibitor in the buffer report different numbers from assays that do not, because the accumulation being measured is a balance between production and breakdown.

The readout itself is the second large lever. A proximal measurement of cyclic AMP accumulation sits close to the receptor. A reporter gene driven by a cyclic-AMP-responsive element sits several amplification steps downstream and integrates signal over hours, which typically left-shifts the apparent potency relative to the proximal measurement. A beta-arrestin recruitment assay reports a physically different event and has no reason to agree with either. Comparing an EC50 from one of these to an EC50 from another is comparing two different experiments that happen to share a unit.

Incubation time interacts with all of the above. Endpoint accumulation assays with a short window sample a rising signal; longer windows approach a plateau and pull the curve leftward. Species variant matters too, since the human receptor and its rodent counterpart are not identical proteins and a peptide optimized against one will not necessarily behave the same at the other. Papers that state only the receptor name and not the species have omitted a variable that can dominate the comparison.

Finally the normalization. If Emax is expressed against a reference agonist, the identity and the concentration series of that reference set the ceiling that everything else is measured against. Two papers using different references produce percentages that are not on the same scale, and the compound has not changed at all between them.

Assay variables and the direction each one moves a reported potency

The consequence for anyone reading across papers is blunt. A table that lists EC50 values from several sources side by side, without listing the cell line, expression system, readout, incubation, species, and reference for each row, is not a comparison. It is a list of numbers that were measured in different experiments and then placed next to each other, and the ranking it appears to show can invert entirely once the conditions are matched. When conditions cannot be matched, the honest move is to say so rather than to average.

Selectivity is a ratio, and a ratio needs its panel

When a molecule engages more than one receptor, the interesting question stops being how potent it is and becomes how its activity divides across the targets. That division is expressed as selectivity, and selectivity is arithmetic performed on measurements rather than a measurement in itself.

The construction is straightforward. Run the same compound against each receptor of interest, fit a curve for each, and take the ratio of the potencies. In log space this is a difference of pEC50 values, which is why selectivity is often quoted in log units. The arithmetic is trivial; everything difficult lives in the word same. For the ratio to mean anything, the arms of the panel have to be matched, and matched is a demanding standard: the same host cell background, comparable receptor expression levels across arms, the same readout, the same incubation, the same buffer, ideally the same experimental session and the same reagent lots.

Break any of those and the ratio picks up an artifact. Suppose one receptor arm is run in a line expressing many copies and another arm in a line expressing few. The first arm is left-shifted for reasons that have nothing to do with the ligand, and the resulting ratio reports the difference between the two cell lines while appearing to report a property of the compound. This is the single most common way a selectivity figure becomes wrong without anyone making an error in any individual experiment.

There is a second decision that is rarely stated. Selectivity computed on potency answers a different question from selectivity computed on efficacy. A compound could be equipotent across two receptors while reaching a much lower ceiling at one of them, which a potency-only ratio reports as no selectivity at all. Some groups combine both into an operational measure of agonist activity so the ratio reflects potency and efficacy together. Those two flavors of selectivity are not interchangeable and a bare ratio does not say which was used.

This is where the word balanced deserves a plain answer. Balanced activity is a design intention, describing what a medicinal chemistry effort was aiming at when the sequence was assembled. It is not a measurement and it has no units. A molecule can be described as balanced by the people who designed it and then measure as clearly skewed in an independent panel, or measure as balanced in one panel and skewed in another with different expression levels. Reading balanced as though it were a result is a category error, and it is one that propagates easily because the word sounds quantitative.

The practical rule follows directly. A selectivity ratio quoted without its panel and conditions is not interpretable, not because it is necessarily wrong but because there is no way to determine whether it is right. Ratios that circulate through secondary writing detached from their source assay should be treated as claims about which paper someone read, not as facts about a molecule.

What a selectivity ratio needs stated before it can be read

A well-reported panel usually looks unimpressive next to a single dramatic ratio, because it comes hedged with conditions and confidence intervals. That is what a defensible measurement looks like. The clean two-digit ratio with no methods attached is the one to distrust, and the fact that it is easier to repeat is precisely why it spreads further than the qualified version it was extracted from.

A reasonable habit when reading any multi-receptor summary is to trace each ratio back one step. If the source is a paper with a methods section, the conditions can be checked and the ratio either survives or does not. If the source is another summary, the trace has not reached anything checkable yet and should continue. Ratios that cannot be traced to a described panel are best recorded as folklore rather than carried forward into a table.

One receptor can report through several pathways

A receptor is not a switch with a single wire behind it. Once activated, a class B G-protein-coupled receptor can couple to heterotrimeric G proteins, principally through the G alpha s route that stimulates adenylyl cyclase and raises cyclic AMP, and it can also be phosphorylated by receptor kinases, recruit beta-arrestin, be internalized, and be either recycled or degraded. Those are physically different events happening on different timescales, and any one of them can be turned into an assay readout.

Biased agonism is the observation that a ligand can favor one of those routes over another relative to how a reference ligand divides its activity. Two compounds with identical cyclic AMP potency can differ substantially in how much beta-arrestin they recruit, and a summary that reports only the cyclic AMP number has erased that difference completely. This is a real limitation of any single-figure potency claim: the figure is not merely incomplete, it is silent about a dimension that exists.

Quantifying bias is harder than observing it. The reason is that different readouts amplify differently, so a compound can look biased simply because one assay in the pair has more amplification than the other. Distinguishing genuine ligand bias from this system bias requires comparing each compound against a reference agonist within each pathway and then comparing those normalized values across pathways, which is the logic behind the operational bias measures in the literature. The practical implication is that a bias claim is always relative to a chosen reference and a chosen pair of assays, and changing either can change the conclusion. A meaningful fraction of reported bias in the broader GPCR literature has later been attributed to assay amplification rather than to the ligand.

There is a further wrinkle specific to peptide agonists. Recruitment and internalization assays often require tagged receptor constructs, and the tags are large. Fusing a fragment of an enzyme or a fluorescent protein to a receptor terminus can alter trafficking and recruitment kinetics, which means the tagged construct is not guaranteed to behave like the untagged receptor. That does not invalidate the assay, but it does mean the arrestin arm and the cyclic AMP arm may be measuring two slightly different receptors.

Time behaves differently across readouts as well. Cyclic AMP accumulation over a short window and receptor internalization measured over a longer one are sampling different phases of the same process, and a compound whose signal persists differs from one whose signal decays quickly even if both reach the same peak. None of that appears in a potency table.

Common readouts at a class B receptor and what each one leaves out

The reason this section belongs in any honest account of multi-receptor pharmacology is that receptor count and pathway count are separate axes. A molecule described by how many receptors it engages has been described along one of them only. Two molecules engaging the same three receptors can divide their activity across downstream pathways very differently, and no arithmetic performed on receptor counts will surface that. Only a panel that reports more than one readout per receptor can, and few published summaries do.

Full and partial agonism depend on the comparator

Calling a compound a full agonist or a partial agonist sounds like a classification of the molecule. It is not. It is a statement about the height of one curve relative to another curve chosen as the reference, measured in one system, and every element of that sentence can be varied.

The convention is to run a reference agonist, usually the native peptide for that receptor, define its plateau as 100 percent, and express the test compound's plateau as a fraction of it. A compound reaching the same plateau is called a full agonist. One reaching a clearly lower plateau is called a partial agonist. A compound exceeding the reference plateau is sometimes described as a super-agonist, which usually says more about the reference than about the test compound.

Two things break the classification. The first is the choice of reference. If the compound picked as the reference is itself submaximal in that system, every other compound is scored against a depressed ceiling and the percentages come out high. Swap in a different reference and the same measured curves produce different labels. Papers that state the reference explicitly can be recalculated by a reader; papers that report percentages without naming the reference cannot.

The second is receptor reserve, which is the same expression-level effect that moves potency, acting here on efficacy. In a system with many spare receptors, a ligand with modest intrinsic activity can still drive the downstream response all the way to its ceiling, because the ceiling is set by a downstream component that saturates before receptor occupancy does. That ligand measures as a full agonist. Move it into a system with low receptor density and the same ligand can no longer reach the ceiling, and it measures as a partial agonist. Nothing about the molecule changed. Only the number of receptors did.

This is why the operational framework in receptor pharmacology separates the ligand's intrinsic ability to activate the receptor from the system's capacity to amplify. The ligand-specific parameter is what one would like to compare across papers; the observed Emax is what actually gets reported. When only the observed value is available, the label attached to it is a joint statement about the compound and the cell line, and it should be quoted with the cell line attached.

For multi-receptor peptides this has a specific consequence. A compound can be full at one receptor of a panel and partial at another, and whether that is described as a designed feature or as an artifact of unequal expression across the panel arms depends entirely on how carefully the panel was matched. The same data can support either reading, which is why the methods section matters more than the summary table.

Worked illustration only. Values below are invented for teaching and are not measured data for any real compound.

Every row of that illustration uses the same hypothetical compound. The label changes because the comparator and the system changed, which is the entire point. When a write-up asserts that a molecule is a partial agonist at one of its targets, the useful follow-up question is not whether that is true but against what and in what, and if the source cannot answer, the assertion has not been established. The values shown are placeholders chosen to make the mechanism visible and should not be quoted anywhere as findings.

What receptor-assay data licenses, and what it does not

It is worth being explicit about the boundary, because the most confident sentences written about multi-agonist peptides are usually the ones that have already crossed it without saying so.

A receptor assay establishes that a molecule, applied at known concentrations to a defined cell system expressing a defined receptor, produced a measurable change in a defined downstream readout, with a fitted potency and a ceiling relative to a named reference. That is the whole claim. It is a real and useful claim. It constrains molecular behavior in a dish and it supports comparisons against other molecules measured in the same dish on the same day.

What it does not do is describe anything about a whole organism. The gap between the two is not a small extrapolation, it is a stack of unmeasured variables: whether the molecule reaches the tissue where the receptor is expressed and at what concentration, how quickly it is cleared or degraded, how much of it is bound to plasma proteins and therefore unavailable, what the actual receptor density is in the relevant tissue compared with the engineered line, what other receptors of the same family are present nearby, and what compensatory responses the surrounding network mounts. Every one of those can change the sign of a comparison, not just its size.

Popular writing about multi-receptor peptides crosses this boundary in a predictable way. A ranking of EC50 values is presented, and then the next sentence describes the compounds as more or less powerful in a general sense. Nothing in the assay supports the second sentence. The move usually happens without a hedge, often without the writer noticing, and it survives because the first sentence is genuinely quantitative and lends its credibility to the second.

There is also a boundary in the other direction that matters for anyone handling material. Published receptor data describes the molecule as characterized by whoever ran the assay, using material they prepared or sourced. It says nothing about the contents of any particular vial. A literature EC50 is not a specification, it is not a certificate, and it cannot substitute for lot-specific identity and purity documentation on the material actually in hand. Those are separate evidence chains and they do not cross-support one another.

The honest position, then, is narrow and defensible. Receptor pharmacology tells you how a molecule engages a receptor in a controlled system, and read carefully it tells you a great deal about what to expect from that system. It does not license statements about organisms, and no amount of stacking receptor numbers converts them into one.

Which statements receptor-assay data supports on its own

The last two rows are the ones that carry the practical weight. A receptor number cannot be promoted into an organism-level statement, and it cannot be demoted into a quality-control record either. Keeping those three evidence chains separate, pharmacology in a defined system, evidence in a living system, and analytical documentation of a physical lot, is most of what it takes to read this literature without overclaiming.

The same discipline applies in reverse when reading a product page or a secondary summary. A potency figure quoted next to a catalog listing is describing published work on a molecule, not the container it appears beside, and the two should be evaluated with different questions and different documents.

Questions this measurement layer gets asked

Where to read next

• Retatrutide research overview: a lipidated peptide as material the physical-material side, solubility, adsorption and chromatography
• Retatrutide vs tirzepatide vs semaglutide: a 2026 research comparison how the evidence bases and documentation differ across the three
• Peptide purity versus peptide identity for research labs the separate evidence chain that describes a physical lot
• Mass spectrometry for peptide identity confirmation what a mass match does and does not establish
• Retatrutide, 10 mg research vial lot-specific documentation, research use only
• Third-party lab testing and certificates of analysis