Phage display of peptides: screening, selection and validation

Phage display peptide technology is a research technique that creates a direct link between the peptide presented on the surface of a bacteriophage and the DNA sequence encoding that peptide. Scientists can expose a very large population of phages displaying different peptide sequences to contact with a selected molecular target, recover phage particles that remain bound to that target, and then determine the sequences of the displayed peptides through analysis of their genetic material. The sequences recovered in this way become candidates for further research. However, the mere fact of their selection does not mean that they are already confirmed ligands; they exhibit high selectivity, biological activity, or therapeutic potential.

This method is particularly useful because it combines the experimentally observed molecular interaction with genetic information that can later be retrieved and analyzed. At the same time, this advantage creates an important interpretational challenge. Binding to the intended target is not the only factor determining which sequences become more abundant during the phage display experiment. The composition of the initial library, interactions with surfaces, and materials used in the experiment, differences in bacterial multiplication, the predominance of certain phages during propagation, the depth of sequencing, and the way subsequent data analysis is performed can all influence which sequences ultimately appear to be the strongest „winners.”.

In this article, we explain the most important conceptual steps for phage display peptide design and the types of evidence needed to move from an enriched peptide sequence to a scientifically justified binding conclusion. We focus on the principles of screening, selection, sequencing, sources of systematic errors, validation, and interpretation of the evidence. The text describes the fundamentals of the research method rather than a detailed laboratory protocol.

How phage display enables the presentation of peptide sequences

The connection of the presented molecule with its genetic record

A bacteriophage is a virus that infects bacteria. In phage display of DNA-encoded peptides, the peptide can be introduced into the system in such a way that the peptide is bound to the phage coat protein. The resulting phage particle carries two related types of information: its genetic material contains the peptide-encoding sequence, while the corresponding peptide is physically presented on the surface of the particle.

This relationship is often described as a genotype–phenotype association. The genotype inside the phage remains associated with the molecular phenotype presented on its surface.

Smith’s groundbreaking 1985 work experimentally demonstrated that foreign sequences can be presented on the surface of filamentous bacteriophages in an antibody-accessible form, and phages containing the recognized sequence can subsequently be enriched. This study laid important conceptual and experimental foundations for the development of modern phage display libraries. However, it should not be interpreted as evidence that any introduced peptide sequence will be expressed, presented, folded, or selected with equal efficiency. [1]

The ability to retrieve genetic information is one of the distinguishing features of phage display compared to the conventional exposure of a target to a complex mixture of unidentified free peptides. After the phage has been recovered following the selection step, sequencing of its genetic material allows the identity of the peptide candidate to be determined.

However, simply knowing the sequence tells primarily what peptide was presented by the recovered phage. It does not automatically explain why exactly this molecule was captured during the experiment. Demonstrating that the peptide presented is indeed responsible for the observed interaction with the target requires additional controls and validation.

The Fag is part of an experimental presentation system

The peptide presented during selection does not exist as a completely independent molecule. It is associated with a much larger biological molecule, and its orientation, availability, local molecular environment, number of copies, and position relative to neighboring coat proteins can influence its interaction with the target.

For this reason, a peptide that appears to strongly bind to a target when presented on the surface of a phage may behave differently after chemical synthesis and analysis as an independent molecule.

This has a direct connection with the difference between affinity and avoidance. Affinity describes the strength of a single molecular interaction between binding partners. Avoidance, on the other hand, refers to the total force that arises when several interactions occur simultaneously.

If several copies of the presented peptide interact simultaneously with many target molecules located on the surface, the phage can be inhibited very effectively thanks to multipoint interactions, even if the affinity of a single isolated peptide molecule is much weaker.

Therefore, strong binding of the phage particle should not be automatically interpreted as evidence of equally strong binding of the free peptide. In a valid scientific description, the presentation system and, if relevant, the molecular form used during subsequent validation should be specified. The results obtained for the isolated synthetic peptide, the entire phage particle, or the same sequence attached to another carrier correspond to related but nevertheless different research questions.

Linear and conformally constrained libraries study other structures

Phage display peptide libraries can also differ in the structures available for the sequences presented.

Linear libraries present sequences without the intentional introduction of constraints that force a loop structure. In conformally constrained libraries, the range of possible peptide conformations is intentionally narrowed, for example by placing cysteine residues on both sides of the random region. The formation of a disulfide bridge can then lead to the creation of a loop.

None of these variants is inherently better in every experiment. The most appropriate architecture depends on the intended goal, the type of molecular interaction being analyzed, and the final form in which the peptide is to be studied.

Names such as Ph.D.-12 and Ph.D.-C7C, often encountered in publications on phage display, refer to different library architectures. The manufacturer's documentation describes Ph.D.-12 as a library containing random peptide inserts of 12 amino acid residues in length, whereas the C7C design contains seven random amino acid positions placed between the cysteine residues that can restrict the structure by creating a loop stabilized by a disulfide bridge.

Importantly, the seven random positions should not be interpreted as all the amino acids present in the entire structure presented. These names describe specific research library designs. They do not guarantee that any of the libraries will lead to the generation of a highly selective candidate for a specific target.[2,3]

The design of the library and its diversity

The library is a population, not just a list of possible sequences

The phage display library begins conceptually with the design of variable DNA sequences and the way they are combined with a phage display system. However, after its physical creation, it becomes a biological population, not merely a mathematical list of all theoretically possible sequences.

The individual inserts must actually be found in the library, be maintained, undergo expression, be properly incorporated into phage particles, and propagate in a living biological system.

For this reason, several different concepts regarding the „size” or „diversity” of a library should not be used interchangeably.

If in a typical library seven entries can be occupied by any of the 20 standard amino acids, the mathematical space of sequences is 20 to the seventh power. Extending the variable region to 12 entries greatly increases this theoretical space.

These calculations describe the number of sequences that could theoretically exist. They do not prove that every possible sequence was actually created, survived the library construction, or exists in a similar number of copies.

This difference becomes increasingly important as the theoretical space of sequences increases. A library can contain an enormous number of physical particles, yet it can only encompass a small fraction of all the mathematically possible peptides.

Randomness does not mean equal representation

The term „randomized” describes the way in which diversity of sequences is created. It does not mean that every amino acid or every possible sequence occurs with the same frequency.

The DNA itself can already introduce an imbalance. The genetic code is degenerate, which means that several different codons can encode the same amino acid, and some codons serve as stop signals. Depending on the way nucleotides are randomly arranged, certain amino acids may be overrepresented even before biological selection begins.

Further deviations may occur during the construction of the library, the expression, the influence of the sequence on the lifespan of the phages and their multiplication in bacteria.

An analysis using next-generation sequencing published in 2022 showed an uneven sequence representation in the 12-amino acid commercial library, the presence of STOP codons, and a significant number of wild-type clones even before the targeted selection was performed. [4]

This does not mean that every commercial library will have exactly the same characteristics. However, the study experimentally shows that an initial library can already possess significant structure and specific sources of bias before it even encounters the target it is being tested against.

The presence of wild-type or non-expected insert-containing particles is particularly important for the interpretation of the number of phage particles. A huge total number of phages does not necessarily mean a similarly large number of unique and properly presented peptide variants.

The place of presentation can influence the bias

The very architecture of the presentation system can influence which sequences will survive and be represented in the library.

In a 2018 study, genetically encoded peptide libraries were tracked at successive stages of construction, biological expression, and subsequent chemical modifications. Significant loss of diversity was observed during biological expression, and differences in the preferences of specific amino acids were observed depending on the location of peptide presentation in the system.[5]

This constitutes experimental evidence that the system's architecture is not merely a passive carrier. It can actively shape the composition of the population available for later selection.

This does not mean that a library containing a bias is worthless. Practically every biological screening system has certain limitations. It is crucial to take them into account when interpreting the results.

A potentially valuable peptide may not appear in the population after selection because it was poorly represented from the start, was presented with low efficiency, or its phage did not replicate well; not because the peptide was unable to bind its target.

Conversely, a sequence that occurs exceptionally frequently at the beginning may remain dominant after selection, partly thanks to its initial numerical advantage.

The individual parts of the library do not have to be identical

Even libraries produced or sold under the same name should not be automatically considered identical in terms of sequence composition.

A 2025 study compared two batches of the same phage display library and tracked changes in their composition during subsequent stages of replication. The researchers found differences both in the initial sequence composition and in the way the populations changed during propagation. Among the strongly enriched candidates were sequences whose predominance could be related to better replication rather than to intended interaction with the target. [6]

The proposal is important for the reproducibility of studies. Simply stating the trade name of the library does not always describe all the characteristics of the initial population. The batch number, its initial composition, and the history of propagation can help explain why two seemingly similar experiments lead to the generation of different candidates.

Selection of binding and enrichment

Selection separates the captured particles from the initial population.

During typical binding-directed selection, the phage library is exposed to a preparation containing a molecular target of interest to the researchers. The material that does not remain bound to the experimental system is removed, while the retained phage particles are recovered.

The recovered population can then be analyzed, multiplied, or used in subsequent rounds of selection. The overall process of this type is often referred to as biopanning.

Although the basic idea may resemble a direct test of peptide binding to a target, each step of the experiment can influence which phagocytes ultimately survive.

The phage may be inhibited because the presented peptide recognizes the intended target. However, it may also interact with the auxiliary surface, the target presentation system, the marker, the uptake reagent, the carrier, or another component of the experimental system.

The process of recovery and subsequent multiplication can introduce further selective pressures.

In practice, the selection experiment therefore selects phage particles based on their ability to undergo the entire experimental process, and not solely on the basis of a single idealized peptide–target interaction. Appropriate controls are necessary to determine which portion of the observed enrichment can actually result from the molecular interaction being studied.

Negative selection tests what else the candidate might recognize.

Counterselection, or negative selection, is used to study or limit interactions with materials that the candidate should not recognize.

Depending on the experiment, these may include a goal-free substrate, an irrelevant protein, a carrier, an uptake reagent, a related cell population, or another element of the experimental system.

The most informative controls depend on the way the target is presented. If the purified target has been immobilized using a specific molecular marker, the possibility of binding with this marker or an uptake system should be considered.

If the objective research question is the identification of a specific type of cell, comparison with similar cells can be more informative than a control that only includes the isolated protein.

The absence of binding to a single control object does not, however, mean universal specificity. Selectivity always has a comparative nature. The statement that a peptide is „specific” remains incomplete until it is known against which alternative targets, surfaces, cells, proteins, or components of the system the test was performed and under what conditions.

Negative selection can limit one type of undesirable interactions, but it cannot rule out all possible effects beyond the target.

Enrichment means a change in the composition of the population

In phage display enrichment means that a given sequence or group of similar sequences becomes more abundant compared to the previous population or the corresponding control.

This is very useful when selecting candidates for further analysis, because it indicates that certain elements of the library are preferentially represented during the experimental process.

However, enrichment is not a direct measure of affinity, constant dissociation, biological activity, or universal selectivity towards the target.

This problem is well illustrated by a study of peptide sequences repeatedly selected in phage display experiments. Ultimately, it was shown that the observed clone success resulted from mutations in another part of the phage genome that enabled more efficient replication even without the intended selective pressure associated with the target. Reconstruction of the corresponding genetic changes allowed the sequence of the presented peptide to be separated from the clonal propagation advantage. [7]

This example shows why repeatedly recovering the same sequence should not be automatically treated as a repeated confirmation of its binding.

Comparing the population after selection with controls that include self-reproduction or other appropriate controls can help determine whether a given sequence becomes more numerous precisely because of contact with the target or because the clone associated with it is more successful during biological propagation.

A higher number of rounds does not automatically mean stronger evidence

Repeating rounds of selection can increase the participation of truly valuable candidates, but it can also reinforce biases already present in the initial population or promote clones that have an advantage in reproduction.

Gradual narrowing of sequence diversity may therefore indicate a useful goal-dependent enrichment, but the mere narrowing of the population does not prove that the molecular quality or affinity of the candidates increases.

If, at an early stage of the experiment, binding to the wrong surface or reagent is preferred, subsequent rounds can further enhance this unwanted interaction more effectively.

Therefore, there is no universal principle according to which a higher number of selection rounds automatically leads to a better result. The appropriate moment for ending the process depends on the research question, the changes observed in the population, the results of the controls, the diversity, and evidence indicating that there is indeed goal-dependent selection.

Additional rounds will not fix the incorrect way of presenting the goal and will not replace independent validation.

Sequencing of selected candidates

Sequencing of Sanger and NGS show a population from different perspectives

Sequencing the individual recovered clones allows the identification of specific peptide candidates and is still useful when specific clones need to be further characterized.

New-generation sequencing, or NGS, allows the analysis of a much larger population. This enables the acquisition of information about the frequency distribution of sequences, families of similar sequences, diversity, and changes occurring between successive stages of selection.

However, none of these methods provides an image of all the molecules present in the population that is perfectly complete and completely free of errors.

Sequencing depth, amplification artifacts, technical errors, sample preparation, filtering thresholds, and bioinformatics decisions can all influence the observed sequence distribution.

Indeed, a rare sequence can be difficult to distinguish from technical noise, whereas a very numerous but biologically insignificant variant can dominate the dataset if its biological context is overlooked.

A 2023 study comparing different sequencing depths showed that a larger number of reads allowed the detection of additional unique sequences and changed the picture of the library's heterogeneity.[8]

This does not mean that there is a specific number of readings that guarantees a complete characterization. The depth of sequencing should be appropriate to the analytical question.

The method of data processing should be clearly described.

The analysis of the sequence becomes much easier to interpret when the principles for transforming raw data into a list of peptide candidates are presented in a transparent manner.

Researchers may need to determine the expected boundaries of the insert, confirm the correct reading frame, define quality criteria, and determine how to handle atypical insert lengths, STOP codons, incomplete reads, and other atypical sequences.

In the publication, it is also worth distinguishing between raw sequencing reads, reads obtained after qualitative filtering, unique DNA sequences, and unique translated peptide sequences.

This distinction is important because due to the degeneration of the genetic code, several different DNA sequences can encode exactly the same amino acid sequence.

In turn, very similar peptide sequences can come from actually related clones, technical errors in sequencing, or different genetic variants.

Counting nucleotide sequences and counting translated peptide sequences therefore answer different questions.

A transparent analytical process should retain sufficient information for another researcher to understand why specific candidates were retained, rejected, combined, or considered a priority.

The final list of peptide sequences itself, without information about how the source data were processed, is much more difficult to evaluate and reproduce.

Sequential motives are hypotheses regarding recognition.

This motif is a recurring pattern of sequence that occurs among similar peptide candidates.

If several independently enriched peptides contain similar amino acids at specific positions, the common residues may be important for identifying the target being studied.

However, mere repeatability does not prove that a given motif is responsible for the binding mechanism.

This same pattern may arise because certain amino acids were overrepresented in the initial library, certain sequences were more effectively presented or replicated, or the experimental system favored a particular structural property unrelated directly to the intended goal.

Comparison with the initial population and appropriate negative controls can help distinguish between these possibilities.

Independent repetitions of the experiment also provide more information than repeatedly analyzing different batches of the same previously selected population.

An attractive-looking sequence logo thus remains a visual summary of the patterns present, rather than an independent experiment confirming the binding. Its role is to show recurring features, facilitate hypothesis generation, and help in choosing amino acids or families of candidates for further study.

Validation of binding and sources of selective bias

The claim must be validated in stages.

Validation is most convincing when the general statement is broken down into several separate questions.

First, it is possible to check whether the chosen phage interacts with the target. Then it is possible to assess whether the presented peptide is indeed responsible for the interaction, whether the interaction is selective with respect to the appropriate controls, and whether its strength changes in a predictable manner with concentration, and whether the binding itself leads to the expected biological or biochemical effect.

These questions should not be combined into one „positive result”.

This scheme illustrates the degree of evidence. It is not a universal set of mandatory experiments. The appropriate validation method depends on the candidate, the molecular target, the presentation system architecture, the biological question, and the potential application.

An independent study of a synthetic peptide could strengthen the conclusions

A particularly valuable step could be removing the candidate from the phage context and checking whether its properties remain after the molecule has been prepared independently.

In a 2009 study, phage display was used to identify chemically restricted cyclic peptide candidates, and then the selected inhibitor was independently synthesized. Subsequent experiments included evaluation of enzyme inhibition and activity in human plasma outside the body. [9]

Such a transition from selection to independent testing provides significantly more information than the enrichment of the sequence alone, but it still remains a preclinical evidence rather than a confirmation of therapeutic efficacy in patients.

In another study, the sequence characteristics obtained during phage selection were transferred to synthetic hybrid peptides and then protease inhibition was evaluated. [10]

Such experiments help determine whether the functional activity actually remains associated with the separately prepared peptide or depends on the properties of the original phage clone.

The exact molecular form remains important in this case. Removing the disaccharide restriction, changing the peptide terminus, attaching a fluorescent marker, modifying the amino acids, or placing the sequence on a different scaffold can change the conformation and binding properties.

Validation should therefore take into account significant structural differences, rather than assuming that every molecule containing the same basic sequence behaves identically.

Analytical certainty is not the same as the possibility of clinical application

A peptide for which binding to a specific target has been convincingly confirmed can become a valuable research reagent, analytical probe, imaging candidate, molecular recognition tool, or a starting point for further development.

None of these results automatically means that the peptide is a drug.

Moving to clinical application requires answering many questions that phage display alone cannot address. These include molecular stability, exposure, tissue distribution, metabolism, toxicity, immunogenicity, production capabilities, formulation, pharmacokinetics, pharmacodynamics, and ultimately whether targeting the selected target leads to a significant outcome in humans.

A successful exploratory experiment does not therefore determine the appropriate route of administration or a safe method of use in humans. We describe more about why the route of administration is a separate element of the evidence regarding peptides in the material. Peptide injections and other routes of administration.

Similarly, demonstrating that a peptide binds to a protein associated with a specific disease does not prove that a change in the activity of this protein will improve the course of the disease.

The result of a phage display study should therefore be described in accordance with the actual stage of the evidence. One study may reveal an enriched sequence, another a validated ligand, and a third a functional inhibitor. Each of these stages represents valuable scientific progress, provided that the conclusion remains proportional to the available data.

Frequently asked questions

Does the enrichment of the sequence prove high variability?

No. Enrichment means that a given sequence has become more represented during the experimental process. In itself, it does not prove high variability. A clone may increase its abundance due to a goal-independent advantage in reproduction or other procedural feature, an example of which are studies in which apparent enrichment was associated with a phage mutation that increases its effectiveness in propagation.[7] Variability requires comparison with relevant alternative targets, surfaces, reagents, or other components of the system. Enrichment should be best treated as a reason for further investigation of the candidate, rather than as a final confirmation of its selectivity.

Can phage display directly indicate clinical therapy?

No. The phage display can identify molecular candidates that warrant further study, but clinical utility requires a much longer chain of evidence. Even if the selected sequence is independently synthesized and shows inhibitory activity in an ex vivo experiment with human plasma, this does not yet mean safety or therapeutic benefit in patients. Studies of cyclic peptides demonstrate a valuable transition from screening to functional tests, but they remain preclinical studies.[9]

What is the genotype–phenotype relationship?

This is the relationship between the molecular feature present on the surface of the phage and the DNA sequence that encodes it. Since the selected fragment contains the genetic information corresponding to the peptide presented, researchers can recover the phage and, through sequencing, determine the peptide sequence. This relationship was the basis of early experiments with fusion phages.[1] However, the genotype–phenotype relationship shows which sequence was associated with the recovered fragment, and not why the fragment survived selection.

Is the phage display library a mixture of free peptides?

No. The peptides in the phage display library are presented as phage-related structures, not as independent peptides freely dissolved in solution. The phage carrier, orientation, number of copies, and presentation geometry can influence the interaction with the target. If the potential application involves a free peptide, its behavior should be studied in this form, and not inferred solely from phage binding. [1,9]

What is the difference between a linear library and C7C?

The difference relates to the architecture of the presented peptide. The linear library does not intentionally introduce limitations in the form of a disaccharide bridge-stabilized loop. In the C7C architecture, the seven random residues are located between the cysteines, which can restrict the conformation of the peptide by forming a disaccharide bridge. The Ph.D.-12 system, on the other hand, uses random 12-amino acid inserts. [2,3] None of these projects guarantees the acquisition of a better candidate for any given purpose.

Does a higher number of particles mean a higher number of unique peptides?

No. The total number of particles and the number of unique peptide variants are two different parameters. A large population may contain many copies of the same clone, and some phages may not contain the expected insert. Sequencing of an unselected commercial library showed both an uneven composition and the presence of wild-type clones.[4]

Are randomized libraries completely random?

No. Randomization describes the method of generating variability, but not the guarantee of uniform representation of all amino acids and sequences. Bias can arise already at the stage of codon design and then be further modified during the construction, expression, survival, and propagation of phages.[5]

Why might a library party be relevant?

Different parties of the same nominally library can start an experiment with a different sequence distribution and change during replication. In a study comparing two parties, differences in their composition and propagation patterns were shown, including the enrichment of sequences potentially related to the advantage of replication.[6] This does not mean that each party will be radically different, but it shows why information about the party can be important for reproducibility.

What is a peptide that is not related to its intended purpose?

This is a sequence presented that is recovered for reasons other than the intended peptide–target interaction. It can interact with the substrate, marker, carrier, capturing reagent, or other component of the system. The associated clone may also simply replicate more efficiently.[7] Therefore, multiple sequence retrieval requires appropriate controls before it is determined to be a specific ligand for the target.

What does counter-selection give?

Counterselection helps identify or reduce the recognition of materials with which the candidate should not be associated, such as goal-free substrates, irrelevant proteins, related cells, or other elements of the experiment. It can reduce the important source of false positive results, but it does not remove all possible biases associated with presentation, amplification, or propagation.[10]

Is the most abundant sequence always the best ligand?

No. The final number of sequences in a population depends not only on binding strength; it can be influenced by initial representation in the library, the ability of clones to survive subsequent stages, and the efficiency of replication. Studies comparing populations of libraries and serial amplification show that the population composition can change significantly without a simple relationship with target affinity. [6]

Does sequencing measure affinity?

No. Sequencing identifies genetic sequences and allows the estimation of their frequency in the analyzed population. Greater depth can reveal greater diversity and more accurately describe the population structure, but it does not convert read frequencies into a constant dissociation or any other direct parameter of affinity. [8]

Does the common motif of the sequence prove the binding mechanism?

No. The common motif shows a recurring pattern of sequence, but its cause still requires explanation. Similar remnants may indeed participate in the recognition of the target, but they may also reflect the construction bias of the library, expressive preferences, propagational advantage, or other experimental pressures.[5] The motif is therefore a good starting point for hypothesis building, but it does not constitute independent evidence of the mechanism.

Why is a synthetic peptide being studied after phage selection?

Studying an independently synthesized peptide allows one to verify whether the observed interaction or function persists after removing the sequence from its original phage context. This helps distinguish the property of the peptide itself from the effect resulting from the phage architecture, multi-value presentation, or other features of the clone. Studies of synthetic inhibitors show how much additional information such a validation step can provide. [10]

Do loyalty and avoidance mean the same thing?

No. Affinity describes the strength of a single molecular interaction, whereas avidity describes the overall effect of many simultaneous interactions. Therefore, a phage displaying several copies of a peptide can interact with a target surface differently than a single molecule of free peptide. Strong binding of the entire phage does not necessarily indicate a strong affinity of the isolated peptide.[9]

Does a larger number of selection rounds eliminate bias?

No. Subsequent rounds can strengthen the actual target-dependent signal, but they can equally effectively strengthen the unwanted bias. Studies of serial amplification show that sequence composition can change significantly even without targeted selection for the target.[6] The number of rounds should therefore be considered as an element of the experimental design, rather than an indicator of quality in itself.

What should a well-documented phage display experiment include?

The description should provide sufficient information for another researcher to understand which population was analyzed, how the goal was presented, what alternative explanations were controlled, how the obtained sequences were analyzed, and how the final candidates were validated.

In practice, this means clearly defining the library and the appropriate batch, the presentation architecture, the way the goal is presented, the control conditions, the main stages of selection, the sequencing method, the principles of sequence processing, and the evidence used to validate the candidates.

Studies of library bias and sequencing depth show why the interpretation of a selected population depends on how it was created and analyzed.[4,8] Clear reporting allows distinguishing a repeatable molecular discovery from a result strongly dependent on the unstated properties of a particular experimental system.

Conclusions

Phage display peptide is an advanced method of molecule discovery because it combines physical selection with information about the sequence that can be recovered. The peptide displayed on the surface of a phage can be exposed to a specific target; the phage can be recovered, and the corresponding genetic sequence can be identified. This linkage between genotype and phenotype enables the study of very large populations of peptides while maintaining the practical possibility of identifying selected candidates.

However, the method requires careful interpretation, because the sequence that ultimately becomes the most dominant does not necessarily have to be the sequence with the highest intrinsic affinity for the intended target.

The outcome may be influenced by the composition of the initial library, the presentation architecture, experimental surfaces, the system for presenting the target, the multiplication of bacteria, the growth advantage of specific clones, the depth of sequencing, and the way of bioinformatics analysis.

For this reason, the enrichment should be considered as the beginning of the candidate’s validation, not its end.

The sequence, the participation of which repeatedly increases during selection, may be a valuable candidate for further study, but stronger conclusions require experiments that allow distinguishing between real binding with the target and alternative explanations for enrichment.

The most convincing phage display studies combine well-characterized starting libraries with appropriate target and negative controls, transparent analysis of changes in the sequence population, and an independent study of the most promising candidates.

If the intended application involves a free peptide, it is particularly important to examine it outside its original phage context. This allows the properties of the molecule itself to be separated from avidity and other effects resulting from its presentation on the phage surface.

In later stages of research, the stability of the peptide and its formulation may also be important. However, this is a separate research question that phage display does not resolve. This issue is discussed in more detail in the article. Storage of peptides – temperature, humidity and the effect of freezing.

Evidence should be interpreted in stages. Identifying an enriched sequence is not the same as demonstrating a confirmed ligand. Confirmation of binding is not the same as demonstrating selectivity. Measurement of affinity differs from demonstrating functional biological activity. Even confirmed biological activity remains something other than demonstrating clinical utility.

The phage display of peptides provides the most scientific value when each proposal remains consistent with the level of evidence actually obtained.

The enriched sequence is a candidate.

Confirmed interaction requires independent evidence.

A functional peptide requires an appropriate functional experiment.

The potential clinical usefulness, if it is even considered at all, requires a completely separate development process and appropriate human studies.

Disclaimer

This article is intended for educational and scientific information purposes only. It explains general issues related to phage display of peptides, including the diversity of libraries, molecular selection, sequence enrichment, sequencing, experimental bias, binding validation, and interpretation of scientific evidence. It does not constitute a laboratory protocol or a step-by-step instruction for performing the experiment.

Discussions of phage display libraries, screening systems, sequence enrichment, binding measurements, and independently synthesized peptides should not be interpreted as instructions for the production, administration, or use of peptides in humans. The identification of sequences using phage display does not mean that a given peptide is safe, biologically effective, therapeutically useful, or suitable for human use.

Similarly, binding to a target does not necessarily mean clinical benefit. A peptide can selectively bind to a specific molecular target without causing a useful biological effect. Conversely, a peptide that exhibits activity in a biochemical system, cell, or in an ex vivo experiment may not exhibit the same activity in the entire organism or in humans.

Clinical development requires evidence that goes far beyond phage display screening. This includes, among other things, appropriate identity and molecular purity characterization, pharmacological properties, stability, exposure, distribution, toxicity, immunogenicity, production quality, and—if appropriate—appropriately designed human studies.

The methods and experimental examples described in this article should therefore be interpreted solely in relation to the systems in which they were actually tested. The results obtained for a single library, phage architecture, how to present the target, the peptide backbone, or the experimental model should not be automatically applied to other systems.

References

1. Smith, G. P. (1985). Filamentous fusion phage: Novel expression vectors that display cloned antigens on the virion surface. Science, 228(4705), 1315–1317. https://doi.org/10.1126/science.4001944
2. New England Biolabs. (n.d.). Ph.D.-12 phage display peptide library. Retrieved September 25, 2026, from https://www.neb.com/de-de/products/e8111-phd-12-phage-display-peptide-library
3. New England Biolabs. (n.d.). Ph.D.-C7C phage display peptide library kit v2. Retrieved September 25, 2026, from https://www.neb.com/en/products/e8212-phd-c7c-phage-display-peptide-library-kit-v2
4. Sloth, A. B., Bakhshinejad, B., Jensen, M., Stavnsbjerg, C., Liisberg, M. B., Rossing, M., & Kjaer, A. (2022). Analysis of compositional bias in a commercial phage display peptide library by next-generation sequencing. Viruses, 14(11), Article 2402. https://doi.org/10.3390/v14112402
5. He, B., Tjhung, K. F., Bennett, N. J., Chou, Y., Rau, A., Huang, J., & Derda, R. (2018). Compositional bias in naïve and chemically-modified phage-displayed libraries uncovered by paired-end deep sequencing. Scientific Reports, 8(1), Article 1214. https://doi.org/10.1038/s41598-018-19439-2
6. Sinkjaer, A. W., Sloth, A. B., Andersen, A. O., Jensen, M., Bakhshinejad, B., & Kjaer, A. (2025). A comparative analysis of sequence composition in different lots of a phage display peptide library during amplification. Virology Journal, 22(1), Article 24. https://doi.org/10.1186/s12985-024-02600-x
7. Brammer, L. A., Bolduc, B., Kass, J. L., Felice, K. M., Noren, C. J., & Hall, M. F. (2008). A target-unrelated peptide in an M13 phage display library traced to an advantageous mutation in the gene II ribosome-binding site. Analytical Biochemistry, 373(1), 88–98. https://doi.org/10.1016/j.ab.2007.10.015
8. Sloth, A. B., Bakhshinejad, B., Stavnsbjerg, C., Rossing, M., & Kjaer, A. (2023). Depth of sequencing plays a determining role in the characterization of phage display peptide libraries by NGS. International Journal of Molecular Sciences, 24(6), Article 5396. https://doi.org/10.3390/ijms24065396
9. Heinis, C., Rutherford, T., Freund, S., & Winter, G. (2009). Phage-encoded combinatorial chemical libraries based on bicyclic peptides. Nature Chemical Biology, 5(7), 502–507. https://doi.org/10.1038/nchembio.184
10. de Marco, R., Azzolini, S. S., Lovato, D. V., Torquato, R. J., Amino, R., de Miranda, A., & Tanaka, A. S. (2010). Validation of phage display method for protease inhibitor selection using synthetic hybrid peptides. Combinatorial Chemistry & High Throughput Screening, 13(9), 829–835. https://doi.org/10.2174/138620710792927394