Peptide Selectivity vs Specificity: How to Read a Target Panel

Peptide selectivity is a relative preference demonstrated across specified targets or readouts under specified conditions. It is not proof that a molecule interacts with only one target. A defensible statement identifies the comparison panel, endpoint, species, experimental system and uncertainty. “Specific” should also be defined rather than treated as a universal guarantee of exclusive action.

First ask what kind of preference was actually measured

A paper may use “selective” to discuss binding to one receptor rather than another, a functional response in one system rather than another, or signalling through one pathway rather than another at the same receptor. Those are related questions, but their evidence is not interchangeable.

IUPHAR’s pharmacodynamics teaching resource describes receptor selectivity as relative rather than equivalent to specificity. That distinction is a useful starting point: the comparison and its experimental scope should remain attached to the adjective. It does not establish a universal numerical threshold for every laboratory question. [1]

Different selectivity questions need different comparisons
Question	Relevant comparison	What it cannot establish alone
Binding preference	Comparable affinity measurements across defined targets	Maximum functional response or a downstream mechanism
Functional preference	Defined responses across specified target systems	That the difference is entirely due to binding affinity
Pathway preference	Readouts at the same receptor, with an appropriate reference	Exclusive recognition of that receptor among all targets
Phenotypic effect	A measured change in a cell or other model	Which target caused the change without supporting evidence

A useful reading habit is to replace the adjective with a question: selective for what, compared with what, measured how? If the methods do not supply those answers, the conclusion should remain narrow. A compound’s familiar name cannot supply missing comparison experiments.

A target panel defines the territory that was examined

The GPCR characterisation chapter of the Assay Guidance Manual discusses receptor-subtype selectivity, other related GPCRs and species-dependent characterisation as distinct parts of a pharmacological profile. A single on-target result is not the same deliverable as a comparison across targets. [2]

Imagine that a study tested targets A, B and C but did not examine D. Its evidence can describe the measured A/B/C comparison. It cannot turn D into a negative result. This sounds obvious, yet summary tables often fill an untested cell with a dash that readers later interpret as “no interaction.”

Use explicit categories such as measured value, supported bound, no activity detected within the stated test range, and not tested. They have different meanings. A result below an assay’s detection capability is not proof of a mathematically exact zero, and an untested target provides no experimental negative at all.

Panel selection also affects interpretation. Comparing only a target of interest with one distant alternative answers a narrower question than examining a relevant set of related subtypes. A large panel is not automatically the best panel: relevance, assay quality and comparability matter alongside the number of rows.

Worked example: calculate a relative binding preference

Hypothetical panel, not measured product data: suppose a report provides the following Ki values from appropriate, comparable binding analyses. The endpoint, molecular entity, species and reporting basis are matched. The target labels are placeholders.

Invented binding panel used only to explain interpretation
Target	Reported Ki	Comparison with A	Status
A	2 nM	Reference	Point estimate
B	200 nM	200 / 2 = 100	Point estimate
C	>2,000 nM	>2,000 / 2 = >1,000	Assumed supported lower bound
D	Not tested	Cannot calculate	No experimental comparison

Under these assumptions, the B-to-A Ki ratio is 100. Because a smaller Ki indicates greater affinity under the binding model, the point estimates describe a 100-fold binding preference for A over B. Define the direction: reversing the division gives 0.01, not a contradictory experiment.

For C, the arithmetic gives a bound, not an exact 1,000-fold result. This teaching row assumes that the analysis genuinely supports the reported Ki bound. A single concentration with no observed response does not automatically justify such a Ki statement. D has no ratio at all because it was not tested.

The Assay Guidance Manual’s receptor-binding chapter explains the modelling distinction between a competition IC50 and a calculated Ki. If a real table provides one endpoint for A and a different endpoint for B, that difference must be resolved before treating their quotient as a binding-selectivity ratio. [3]

Logarithmic values need a different arithmetic step

Another hypothetical presentation uses pKi, defined here as −log₁₀(Ki expressed in mol/L). For the invented panel, 2 nM corresponds to about 8.70 and 200 nM to about 6.70. Subtracting gives a difference of approximately 2 log units. The corresponding affinity ratio is 10² = 100, not a two-fold preference.

This is a common place for spreadsheet mistakes. Check whether a field contains a concentration, a logarithm of a concentration, a percentage response or an already calculated ratio. The same numerical operation cannot be applied to every column merely because the entries are numbers.

Preserve the original endpoint and units alongside any transformed value. If a database normalises the presentation, retain enough information to reconstruct the transformation. Our affinity-versus-potency guide covers the different endpoint definitions; the additional question here is whether a comparison across targets is valid.

Point estimates do not remove uncertainty

The ratio in the example uses deliberately clean values to show the arithmetic. Real estimates have experimental variability and modelling uncertainty. Reporting many decimal places after dividing two uncertain values does not make the ratio precise.

Consider what would happen if repeated measurements moved the estimates closer together. The apparent strength of the preference would change even though the labels A and B had not. A useful comparison therefore retains the uncertainty information provided by the source and the basis of replication rather than carrying only rounded midpoints into a ranking.

The data-standardisation chapter of the Assay Guidance Manual discusses result definitions, normalisation and curve processing. Those details are part of what makes a result interpretable, not administrative decoration to discard after extracting a headline value. [4]

No universal cut-off in this article separates “selective” from “not selective.” The required evidence depends on the research question, the alternative targets, the system and the uncertainty. A stated ratio is more useful than a free-standing adjective because readers can assess that context for themselves.

Species, constructs and readouts can change the comparison

A target name is not a complete experimental description. Species, receptor construct, cellular background and readout can differ between studies. The GPCR guidance specifically addresses how the experimental system, including receptor expression, affects observed pharmacological properties. [2]

Consequently, an affinity result at one species’ receptor and a functional midpoint at another species’ receptor should not be combined into a neat “selectivity index” without explaining the mismatch. Such a number might be arithmetically possible while answering no well-defined biological question.

For literature extraction, create separate fields for target identity, species, construct or preparation, endpoint, response definition, reference compound and timing where reported. Compare rows with compatible meanings first. Keep cross-study observations explicitly descriptive if the conditions cannot support a controlled ranking.

This does not make different systems uninformative. A changed profile across models may be an important finding. It simply needs to be described as a model-dependent observation, with the unresolved causes left visible. The research-systems guide explains why moving between models adds another inference step.

Pathway bias is not the same as target exclusivity

The IUPHAR community guidelines for GPCR ligand bias define comparisons in relation to a reference ligand, pathways, receptor and experimental system. A preference between two signalling readouts at one receptor is therefore a different claim from binding preference between two receptor subtypes. [5]

For example, observing a stronger relative signal in pathway P than pathway Q does not establish that the molecule fails to interact with targets B or C. The experiment did not ask that question. Conversely, a binding panel showing a preference for receptor A does not reveal how every downstream pathway at A behaves.

When a summary uses “functionally selective,” read the author’s definition and reference system. Do not silently translate it into “acts only at one receptor.” Keeping the comparison axis visible prevents an interesting signalling observation from becoming a much broader unsupported mechanism claim.

Check alternative explanations for an apparently clean profile

A signal can depend on the assay technology as well as on the proposed biology. The Assay Guidance Manual’s high-content screening chapter describes interference and artifacts that can complicate interpretation of cellular readouts. Its examples do not diagnose any particular peptide experiment; they show why appropriate controls and complementary evidence matter. [6]

Ask whether the measured response depends on the intended target, whether the comparison system can produce the expected response, and whether the material and vehicle are comparable across conditions. A negative result from a poorly functioning alternative-target assay is weak evidence for preference. An independent readout may address a detection artifact, but only if it does not merely reproduce the same vulnerability.

Material handling can add another complication. If the amount reaching an assay differs between conditions, nominally matching concentrations may not represent matching exposure in the tested liquid. Our solubility, recovery and stability guide explains why a lower measured signal needs investigation rather than an automatic mechanism label.

A compact way to report the conclusion

For the invented panel, a defensible summary would identify a Ki-based preference for A over B under matched conditions, retain C as a supported lower-bound comparison and state that D was not tested. It would not describe the molecule as universally specific, convert the ratio into a therapeutic advantage, or imply that unexamined targets had been excluded.

When the source is incomplete, say exactly what is missing: the comparator, compatible endpoint, experimental system, uncertainty or panel coverage. This makes the summary useful for selecting the next evidence question instead of creating a stronger-sounding conclusion than the data support.

Common questions

Does selective mean that there are no other targets?

No. It describes a relative preference within a defined comparison. Unexamined targets and untested conditions remain outside the evidence.

Can I divide an EC50 by a Ki to obtain selectivity?

Not as a straightforward binding- or functional-selectivity ratio. The endpoints describe different questions; compatibility and any model connecting them must be established first.

Does no activity detected mean zero affinity?

No. Detection capability, tested range and the assay model constrain the interpretation. Report the actual observation or supported bound, not an exact zero.

Does a two-unit pKi difference mean two-fold selectivity?

No. Under the stated base-10 definition, two log units correspond to a 100-fold affinity ratio. State the direction and preserve the original units.

This is an interpretation guide with original hypothetical calculations, not a comparative study of catalog materials or evidence of personal-use suitability. See experimental controls and replication and the peptide science library. HealthyWelness operates a commercial research-material catalog; materials are for laboratory research, not human or veterinary use.

Cover: AI-generated conceptual editorial artwork, not an exact molecular structure, measured result or a photograph of a HealthyWelness facility.