How Protein–Peptide Interactions Shape Cell Function

Cells are run by committees of proteins that talk to each other through short peptide motifs. Disrupting or mimicking that conversation is how modern therapeutics work.

Protein–peptide interactions shape cellular function through one conserved structural principle: a short linear motif (SLiM) of 3–10 residues docking into a modular recognition domain. This architecture coordinates signalling cascades, organelle trafficking, chromatin regulation, and immune recognition with the same atomic logic. A peptide, operationally defined as a chain of up to approximately 50 amino acid residues, carries sufficient chemical information in a compact footprint to recruit, displace, or allosterically remodel its binding partner.

This convergence is not accidental. Because the same SLiM–domain architecture recurs across biology's most consequential regulatory systems, disrupting or mimicking a single short sequence can propagate effects through an entire pathway. That mechanistic leverage is why peptide interfaces have become a dominant design focus in cell-biology-derived drug discovery, spanning stapled peptides, peptidomimetics, and TCR-engineered cell therapies currently advancing through clinical trials in Europe.

This article covers how the same structural principle operates across four distinct cellular contexts; why that convergence makes peptide interfaces a primary target for drug design; and how structural genomics and computational tools now enable rational design of peptide-interface inhibitors.

What Are Protein–Peptide Interactions? A Working Definition

A protein–peptide interaction is the non-covalent, reversible binding of a short linear sequence within a peptide chain to a structured recognition domain on a partner protein. This molecular event mediates signal transduction, protein trafficking, and chromatin regulation across virtually every cellular context. Peptides are chains of amino acid residues joined by covalent peptide bonds formed through condensation reactions; by the working convention used throughout this article, that chain spans 2–50 residues, following the operational definition described in Forbes and StatPearls (NCBI, 2023). Some literature extends the upper boundary to 100 residues, particularly for secreted bioactive peptides, but no 2024–2026 IUPAC revision has formalised a new threshold. Any chain exceeding 50 residues is treated here as a protein domain or intrinsically disordered region rather than a peptide.

Why Reversibility and Modularity Define the Biology

Reversibility is not a biochemical footnote; it is the property that makes peptide-mediated regulation possible at all. Because the interaction is non-covalent, a single recognition domain can cycle through multiple binding partners in response to post-translational modifications, concentration gradients, or competing ligands, without requiring new protein synthesis. London et al. (Cell, 2010), a study cited more than 500 times, established that protein–peptide interactions of precisely this reversible character mediate both signal transduction and protein trafficking, two processes that demand rapid, switchable molecular decisions rather than permanent associations.

Modularity compounds this advantage. A recognition domain such as an SH3, PDZ, or bromodomain retains its binding geometry regardless of the larger protein in which it is embedded, meaning evolution can redistribute the same docking surface across unrelated scaffolds. The JCSG Structure Gallery documents atomic-resolution examples of this portability across hundreds of domain–peptide complexes deposited in the PDB.

The Unifying Structural Principle

The mechanistic thread running through every section of this article is a single observation: one SLiM of 3–10 residues docks into one modular recognition domain, and this geometry recurs without fundamental variation across signalling cascades, organelle trafficking, chromatin regulation, and immune recognition. For a deeper account of how that docking geometry is measured and predicted, see Understanding Protein–Peptide Interactions. The fact that so many regulatory systems converge on one structural solution is the reason peptide interfaces have become a primary design target in modern therapeutics, a point developed in each mechanistic section that follows.

The Unifying Structural Principle: SLiMs and Modular Domains

A short linear motif (SLiM) is a stretch of 3–10 amino acid residues, typically embedded within an intrinsically disordered region, that adopts a defined secondary structure only upon engaging its cognate recognition domain. This bind-on-contact mechanism is the single structural logic underlying signalling, trafficking, chromatin regulation, and immune recognition alike.

What Makes a SLiM Functional

The residues that confer specificity within a SLiM are sparse: usually two or three anchoring side chains that insert into pre-formed pockets on the recognition domain, with the intervening positions tolerating considerable sequence variation. This degeneracy is not a flaw; it is the mechanism by which one domain family can recognise hundreds of distinct sequence contexts across the proteome. Johansson-Åkhe et al. (2019), cited 86 times, mapped protein–peptide interaction sites associated with human disease and demonstrated that pathogenic mutations cluster disproportionately at these sparse anchor positions, confirming that a small number of contacts carry the full energetic weight of recognition.

The Architecture of Modular Recognition Domains

Each recognition domain (SH2, PTB, SH3, PDZ, WW, bromodomain, chromodomain, BIR) is a structurally conserved fold whose α-helices and β-sheets define a binding groove of fixed geometry. The groove does not change when the domain is transplanted into a new protein context, which is why evolution has redistributed identical docking surfaces across unrelated scaffolds. Atomic-resolution examples of this portability across hundreds of domain–peptide complexes are catalogued in the JCSG Structure Gallery, drawing on PDB depositions that span all four cellular contexts discussed in this article.

Affinity, Reversibility, and Therapeutic Tractability

Dissociation constants for SLiM–domain interactions typically fall in the micromolar range, and that modest affinity is the point. London et al. (Cell, 2010), cited more than 500 times, established that this reversible character is what enables both signal transduction and protein trafficking to operate as rapid, switchable decisions rather than permanent associations. A covalent bond would be catastrophic in a phosphorylation cascade that must reset within seconds. The same low-affinity, high-specificity geometry that makes SLiMs biologically indispensable also makes them therapeutically tractable: a designed peptide or peptidomimetic can compete for the groove without needing to match the binding energy of an antibody.

Work published in ACS Chemical Biology (28 August 2015) showed that α/β-peptide foldamers, backbone-modified mimetics that resist proteolysis, can target intracellular protein–protein interfaces with sufficient potency to perturb signalling in cell culture, demonstrating that the SLiM–domain groove is accessible to synthetic ligands even inside the cell. For a detailed account of how binding geometry is measured and predicted across domain families, see Understanding Protein–Peptide Interactions. The four functional sections that follow each show the same groove-and-motif logic operating in a distinct cellular context, which is why disrupting it has become a central strategy in modern drug design.

Signalling Cascades: Phosphopeptide Motifs and SH2/PTB Domains

Receptor tyrosine kinases control signal transduction by writing phosphotyrosine marks onto their own cytoplasmic tails, then presenting those marks as structured peptide motifs that downstream adaptors read with exquisite selectivity. The write–read logic is the same groove-and-motif principle described above, but here the modification state of a single residue determines which proteins are recruited, at what stoichiometry, and for how long.

The Write Step: Autophosphorylation Creates Docking Motifs

When a growth-factor receptor such as EGFR or PDGFR dimerises, trans-autophosphorylation on specific tyrosine residues does not simply activate the kinase domain. Each phosphotyrosine (pTyr) residue, together with the three to five flanking residues, constitutes a distinct SLiM that is now competent to engage a modular recognition domain. The motif did not exist as a functional entity before phosphorylation; the post-translational modification creates it. This is why kinase activation and adaptor recruitment are inseparable events rather than sequential ones.

SH2 and PTB Domains: Two Geometries for the Same Mark

SH2 domains recognise pTyr-X-X-hydrophobic motifs, with the pTyr slotting into a positively charged pocket and the +3 hydrophobic residue contacting a second specificity-determining surface. PTB domains bind pTyr-containing motifs in the opposite orientation, reading residues N-terminal to the phosphotyrosine rather than C-terminal. The structural consequence is that a single phosphorylated tail can simultaneously present distinct faces to SH2-containing adaptors (GRB2, SHP2, PI3K p85) and PTB-containing proteins (SHC, IRS-1), allowing branched signalling from one activation event. Which branch dominates depends on the relative affinities and local concentrations of each domain, not on a separate regulatory switch.

GRB2 binding to the EGFR pTyr-1068 motif recruits SOS1 and activates RAS–MAPK, driving proliferative output. PI3K p85 binding to pTyr-1173 activates AKT, promoting survival. The amplitude and duration of each response are set by the affinity of the relevant SH2–phosphopeptide pair, which is why small differences in motif sequence translate into large differences in pathway bias. This is well-established mechanistic ground in receptor biology, though the precise affinity constants for individual SH2–pTyr pairs continue to be refined as structural databases expand.

Computational prediction of which SH2 or PTB domain binds a given phosphopeptide has advanced substantially since the statistical docking approaches published around 2019. AlphaFold-Multimer and related transformer-based pipelines now achieve near-experimental accuracy for well-structured recognition domains including SH2 and PTB families. Benchmarking data for these tools against curated phosphopeptide sets is a fast-moving area that warrants verification against 2025–2026 primary literature before any specific accuracy figure is cited.

Therapeutic Angle: Competing for the Groove

Because SH2 domains bind short, defined peptide sequences with micromolar affinity, they are accessible to competitive inhibitors that do not need antibody-scale binding energy. Peptidomimetics that present a pTyr surrogate (typically a phosphonate or sulfotyrosine analogue) alongside the correct +3 hydrophobic residue can occupy the SH2 groove and block adaptor recruitment. Work reviewed in the context of peptide interface targeting has shown that backbone-modified foldamers can maintain sufficient cell permeability to perturb SH2-dependent signalling in culture, addressing the historical liability of phosphopeptide mimetics being excluded from cells.

The SH2 groove is, in structural terms, a canonical peptide-binding cleft of the type catalogued across domain families in the JCSG Structure Gallery, where atomic-resolution data on domain–peptide complexes illustrate the conserved geometry that makes competitive inhibition feasible. For a broader account of how affinity and selectivity are measured across these domain families, Understanding Protein–Peptide Interactions provides the methodological context. The signalling cascade is, in this sense, a molecular decision tree written in phosphopeptide motifs and read by modular domains, and every node in that tree is, in principle, a target.

Organelle Trafficking: Sorting Signals as Peptide Addresses

Every newly synthesised protein carries its destination encoded in its sequence. Sorting signals are short peptide sequences (typically 5–30 residues, positioned at the N- or C-terminus or within cytoplasmic loops) that are decoded by cytosolic receptor proteins or membrane-bound import machinery, routing the protein to the correct compartment before it can act.

Nuclear Localisation Signals and Importins

The nuclear localisation signal (NLS) is a canonical example of the SLiM-as-address principle. Classical monopartite NLS sequences present a cluster of basic residues (the prototypical SV40 large-T antigen motif PKKKRKV) that dock into the adaptor groove of importin-α, which then recruits importin-β to ferry the cargo through the nuclear pore. The interaction is electrostatically driven: the basic patch on the peptide engages an acidic binding channel on importin-α with low-micromolar affinity, and RanGTP hydrolysis in the nucleus provides the directionality. Mutation of even a single lysine within the NLS is sufficient to strand a transcription factor in the cytoplasm, a mechanism documented in several cancer-associated p53 variants where nuclear exclusion abolishes tumour-suppressive function.

Mitochondrial Targeting Sequences and the TOM/TIM Gate

Mitochondrial targeting sequences (MTS) are amphipathic helices of 15–40 residues at the N-terminus that are recognised by the translocase of the outer membrane (TOM20 subunit) via a hydrophobic groove that accommodates the helical face of the signal. The MTS is cleaved by the mitochondrial processing peptidase after import, making the interaction inherently transient. Mutations that disrupt MTS amphipathicity cause misrouting of metabolic enzymes and are associated with mitochondrial disease phenotypes, including certain presentations of Leigh syndrome.

ER Signal Peptides and the KDEL Retention Code

Signal peptides direct co-translational insertion into the ER membrane via recognition by the signal recognition particle (SRP), which binds the hydrophobic core of the signal through an M-domain methionine-rich groove. Soluble ER-resident proteins carry a C-terminal KDEL tetrapeptide that is recognised by the KDEL receptor on Golgi membranes, triggering retrograde COPI-coated vesicle transport back to the ER. The KDEL–receptor interaction is pH-sensitive: the slightly acidic Golgi lumen promotes binding, while the near-neutral ER pH promotes release. This is a neat example of how the same peptide motif can encode conditional recognition.

Lysosomal Sorting: YXXΦ and Dileucine Motifs

Lysosomal membrane proteins and soluble hydrolases destined for the lysosome carry cytoplasmic tail motifs recognised by the adaptor protein (AP) complexes. The tyrosine-based YXXΦ motif (where Φ is a bulky hydrophobic residue) binds the μ subunit of AP-1 and AP-2 through a defined peptide-binding pocket, directing cargo into clathrin-coated vesicles. Dileucine-based [DE]XXXL[LI] motifs engage the β and σ subunits of AP complexes through a distinct but structurally analogous groove. Loss-of-function mutations in these motifs, or in the AP subunits themselves, cause lysosomal storage disorders: mutations in the YXXΦ motif of LIMP-2 disrupt glucocerebrosidase trafficking and contribute to Gaucher disease pathology, while AP-1 subunit mutations underlie a subset of neurodegeneration-associated lysosomal dysfunction phenotypes.

Atomic-resolution structures of AP–peptide complexes, alongside importin–NLS and TOM20–MTS co-crystal data, are catalogued in the JCSG Structure Gallery, where the conserved geometry of peptide-binding grooves across these unrelated receptor families is directly visible.

Therapeutic Exploitation of Sorting Signals

The programmability of sorting signals has direct therapeutic utility. Gene therapy vectors routinely incorporate engineered NLS sequences to improve nuclear delivery of episomal DNA in non-dividing cells, increasing transduction efficiency in post-mitotic tissues such as neurones and cardiomyocytes. In targeted protein degradation, PROTAC-adjacent strategies append mitochondrial or lysosomal sorting sequences to degrader constructs to concentrate activity at the relevant compartment, reducing off-target cytosolic proteolysis. Separately, synthetic KDEL-tagged toxin fusions exploit ER-retrieval to accumulate cytotoxic payloads within the secretory pathway of tumour cells that overexpress KDEL-receptor isoforms.

The mechanistic logic is identical across all four destinations: a short linear motif presents a defined chemical surface to a modular recognition domain, and the outcome (nuclear import, mitochondrial insertion, ER retention, or lysosomal delivery) is determined entirely by which domain reads it. For the interaction-measurement methods that underpin affinity and specificity data for these sorting complexes, Understanding Protein–Peptide Interactions provides the relevant methodological framework.

Gene Regulation: Chromatin Readers and Peptide-Mimetic Drugs

Histone N-terminal tails are the clearest demonstration in the nucleus that post-translational modifications function as short peptide motifs: acetylated, methylated, or phosphorylated residues within these disordered tails present defined chemical surfaces that modular reader domains decode with high specificity, and the transcriptional outcome depends entirely on which combination of readers is recruited.

The Histone Code as a SLiM-Reading Problem

Each core histone extends an unstructured N-terminal tail of 20–40 residues into the nucleosomal environment. Acetylation of lysine residues within these tails (H3K27ac and H3K9ac being the most studied) creates a hydrophobic, electrostatically neutralised pocket-binding motif recognised by bromodomains. Bromodomains are approximately 110-residue helical bundles whose conserved asparagine residue hydrogen-bonds directly to the acetyl carbonyl, a geometry visible in multiple PDB entries and catalogued in the JCSG Structure Gallery. Chromodomains read a structurally distinct signal: trimethylation of H3K9 or H3K27 generates an aromatic cage that coordinates the trimethylammonium group through cation-π interactions, recruiting HP1 or Polycomb repressor complexes respectively. Phosphorylation of H3S10 during mitosis is read by 14-3-3 proteins via their canonical amphipathic groove, the same groove that reads phosphoserine motifs in the signalling cascades discussed earlier.

The combinatorial logic (often called the histone code) means that a single lysine residue can switch a tail from a bromodomain ligand to a chromodomain ligand simply by changing its modification state. This is not metaphor; it is a direct consequence of the chemical complementarity between a four-residue modified peptide and a pre-formed binding groove, the same structural principle that governs SH3 and PDZ recognition.

BET Bromodomain Inhibitors: Competitive Displacement of an Acetyl-Lysine Motif

BET-family proteins (BRD2, BRD3, BRD4, and BRDT) each carry two tandem bromodomains that bind acetylated histone tails at super-enhancers, concentrating transcriptional machinery at oncogenes including MYC and BCL2. JQ1, the thienodiazepine developed by the Bradner laboratory, competitively occupies the acetyl-lysine binding pocket of BRD4 with a Kd of approximately 50 nM, physically displacing the bromodomain from its histone peptide ligand and collapsing super-enhancer-driven transcription. The mechanism is a direct peptide-interface competition: a small molecule mimics the acetyl-lysine side chain and the flanking backbone contacts that the histone tail makes within the groove.

Clinical successors to JQ1 (including mivebresib, molibresib, and zavabresib) have progressed through Phase I and Phase II trials in haematological malignancies and solid tumours. As of early 2026, the FDA granted orphan drug designation to zavabresib for myelofibrosis, confirming its investigational status. No BET bromodomain inhibitor has received marketing authorisation from the EMA or MHRA for any oncology indication as of 2025–2026; all remain clinical-stage. Researchers using these compounds should verify current trial phases and regulatory status against ClinicalTrials.gov and the EMA clinical trials register before drawing conclusions about approved use.

The broader point is structural: BET inhibitors are among the most clinically advanced examples of a drug class designed explicitly to disrupt a protein–peptide interface rather than an enzyme active site. The mechanistic and affinity data underpinning their development (isothermal titration calorimetry, surface plasmon resonance, and co-crystal structures of bromodomain–acetyl-peptide complexes) are the same measurement frameworks described in Understanding Protein–Peptide Interactions. That the same structural logic operating in cytoplasmic signalling and organelle sorting also governs nuclear gene regulation is not coincidental; it reflects the evolutionary economy of deploying one recognition architecture across every compartment where rapid, reversible, and combinatorially flexible protein recruitment is required.

Immune Recognition: The MHC–Peptide–TCR Handshake

Adaptive immunity operates through a protein–peptide interaction that is structurally analogous to every SLiM–domain complex described above, with one conceptually distinct feature: here the peptide is the antigen itself rather than a regulatory docking motif embedded in a larger protein.

How MHC Molecules Present Peptide to T Cells

MHC class I molecules bind peptides of 8–10 residues derived from intracellular protein degradation via the proteasome; MHC class II molecules accommodate longer fragments of 13–25 residues processed from extracellular or endosomal proteins. In both cases, the peptide sits in a groove formed by two α-helices flanking a β-sheet floor, with anchor residues at defined positions making sequence-specific contacts with polymorphic pockets in the groove. Dissociation constants for stable peptide–MHC complexes typically fall in the nanomolar range, and the complex is kinetically stable enough to persist on the cell surface for hours, giving circulating T cells sufficient time to survey it.

T-cell receptors do not read the peptide in isolation. The TCR contacts both the peptide side chains and the MHC α-helices simultaneously, making the recognition event a ternary interaction: TCR, peptide, and MHC are all structural participants. This dual-contact geometry means that TCR specificity is determined by the combination of peptide sequence and MHC allotype, which is the molecular basis of MHC restriction. The structural logic is well established from decades of crystallography, and atomic-resolution structures of TCR–peptide–MHC complexes are among the most extensively deposited ternary assemblies in the PDB.

Therapeutic Engineering of the Peptide–MHC Interface

Cancer immunotherapy has moved from exploiting this interface passively to engineering it with precision. Neoantigen vaccines present tumour-specific peptides (arising from somatic mutations unique to a patient's cancer) to prime or expand cytotoxic T cells that recognise those peptide–MHC combinations on tumour cells. The therapeutic logic depends entirely on the structural selectivity of the TCR–peptide–MHC handshake: a single amino acid substitution in the peptide can abolish or create TCR recognition, which is why neoantigen prediction pipelines now integrate MHC-binding affinity modelling alongside mutational calling.

TCR-engineered T-cell therapies (TCR-T) take a more direct approach, introducing a high-affinity TCR specific for a defined tumour-associated peptide–MHC complex into a patient's T cells. Targets under clinical investigation include NY-ESO-1 and MAGE-A4 peptides presented on HLA-A*02:01. As of 2025–2026, no TCR-T product has received marketing authorisation from the EMA or MHRA; all remain in early- to mid-stage clinical trials under the EMA's Advanced Therapy Medicinal Product framework. Researchers and clinicians should treat TCR-T as a clinical-stage modality rather than an approved therapeutic option in the UK.

The mechanistic parallel to the rest of this article is precise. A short linear peptide docks into a structured groove on a partner protein; the affinity and specificity of that interaction determine a downstream biological outcome. The difference is that in immune recognition the peptide arrives exogenously, processed from a pathogen or tumour protein, rather than being encoded as a cis-regulatory motif within a signalling scaffold. That distinction is biological context, not structural principle. The same measurement frameworks (co-crystal structures, surface plasmon resonance, isothermal titration calorimetry) used to characterise bromodomain–acetyl-peptide or SH3–PxxP complexes are applied identically to TCR–peptide–MHC assemblies, as the structural data catalogued in the JCSG Structure Gallery illustrates across domain families. Evolution found one solution to the problem of selective, reversible, high-specificity molecular recognition and deployed it from the cytoplasm to the immunological synapse. A fuller account of the binding thermodynamics and affinity measurement methods common to all these systems is in Understanding Protein–Peptide Interactions.

Why Convergence Matters: Peptide Interfaces as the Dominant Drug Target Class

Peptide interfaces are the most tractable class of protein–protein contact for therapeutic intervention because they combine a defined binding groove, a known pharmacophore geometry, and a structural database dense enough to support rational design from first principles. The four cellular systems examined above (signalling cascades, organelle trafficking, chromatin regulation, and immune recognition) all reduce to the same physical transaction: a short linear motif presenting two to four key side chains into a pre-formed hydrophobic or electrostatic pocket on a modular recognition domain. That structural identity is why a medicinal chemist who has solved a BRD4–acetyl-lysine inhibitor problem carries directly transferable knowledge to an SH3–PxxP or PDZ–C-terminal motif problem.

The Transfer of Lessons Across Domains

Every solved structure in one system becomes a template for another. London et al., writing in Cell in 2010 in a study cited 501 times, catalogued the binding strategies used across peptide–protein complexes and showed that a small number of recurring pharmacophore arrangements account for the majority of observed interfaces. The structural vocabulary is therefore finite and learnable. Gupta et al. in 2022 extended this logic to the design of protein fragments and constrained peptides, demonstrating that once the pharmacophore geometry is known, synthetic mimetics (including α/β-peptide foldamers of the type characterised by the ACS in 2015) can reproduce binding affinity while resisting proteolytic degradation. The design cycle is: identify the SLiM, solve the complex, extract the pharmacophore, build the mimetic.

Structural Genomics as the Enabling Infrastructure

None of this is possible without atomic-resolution structures at scale. Structural genomics programmes, including the JCSG pipeline, systematically determined domain structures across protein families, populating the PDB with the recognition-domain templates that underpin structure-based design. The JCSG Structure Gallery illustrates this directly: bromodomains, SH3 domains, PDZ modules, and MHC grooves are all represented, and each entry is a potential starting point for inhibitor design. PDB deposition statistics for 2024–2025 confirm continued growth in peptide–protein complex entries spanning all four cellular contexts covered here.

Computational Acceleration and the 2025–2026 Frontier

Computational identification of cryptic peptide-binding sites has accelerated sharply. The Johansson-Åkhe et al. 2019 statistical docking method established a baseline for proteome-wide SLiM–domain interaction prediction; AlphaFold-Multimer and its community derivatives (ColabFold 2.0, AF-Peptide, and PepFlow) now achieve near-experimental accuracy for well-structured recognition domains and have been benchmarked against curated protein–peptide datasets through 2024–2025. For atomic-resolution prediction of specific complexes, AlphaFold-based multimer models are the current de facto standard; Johansson-Åkhe-style motif scanning retains value for large-scale disordered-region surveys where structural templates are absent. RFdiffusion-derived generative models adapted to peptide interfaces were reported as research-grade in 2024–2025 and are likely to reach routine use within the next two to three years.

The regulatory picture reinforces the strategic argument. As of 2025–2026, no BET bromodomain inhibitor holds EMA or MHRA marketing authorisation, and TCR-T therapies targeting peptide–MHC remain clinical-stage. The pipeline is real and advancing, but the approved-drug density is still low relative to the structural opportunity. The field is earlier in its exploitation of this target class than the mechanistic maturity would suggest. A fuller treatment of the binding thermodynamics that underpin all these systems is in Understanding Protein–Peptide Interactions.

Structural Genomics and the Atomic Maps Behind Drug Design

Structural genomics converted protein structure determination from a craft into a pipeline: high-throughput X-ray crystallography, cryo-EM, and NMR were systematically applied to entire fold families, producing atomic-resolution coordinates for hundreds of recognition domains and their bound peptide ligands in a single coordinated effort. Those coordinates are the foundation on which drug design rests, because a structure at 1.5–2.5 Å resolution resolves the individual hydrogen bonds, van der Waals contacts, and buried water molecules that define selectivity between a SLiM and its cognate domain. Without that atomic map, medicinal chemists are guessing; with it, they can enumerate every contact and ask which ones a small molecule or constrained peptide can replicate.

The JCSG infrastructure illustrates what systematic structural coverage looks like in practice. The JCSG Structure Gallery archives atomic-resolution structures across fold classes, providing experimentally validated templates for peptide-binding pockets that computational models can be tested against. The accompanying JCSG pipeline documents the crystallisation-to-deposition workflow that made high-throughput deposition feasible, and the Protein Sequence Comparative Analysis (PSCA) tool enables researchers to map sequence conservation directly onto those structures, identifying which residues in a binding groove are invariant across paralogues and therefore safe to target without off-target liability.

AlphaFold-Multimer and its derivatives, benchmarked against curated peptide-complex datasets through 2024–2025, have expanded the tractable structural space considerably, generating plausible binding-mode hypotheses for domain–SLiM pairs where no crystal structure exists. That expansion is real, but it does not replace experimental validation: predicted peptide-binding modes for disordered or conformationally flexible interfaces carry uncertainty that only crystallography or cryo-EM can resolve. The productive workflow in 2025–2026 uses AI prediction to prioritise which complexes to crystallise, then deposits the experimental structure to anchor the next round of design. Post-structural-genomics pipelines continue to populate the PDB with SLiM–domain complexes across signalling, trafficking, chromatin, and immune contexts, providing the mechanistic templates that connect atomic detail to the therapeutic arguments made throughout this article. A detailed account of the binding thermodynamics that those structures encode is in Understanding Protein–Peptide Interactions.

Key Terminology: A Consistent Glossary for This Article

A peptide is a chain of amino acid residues linked by peptide bonds; this article follows the StatPearls/NCBI operational convention of 2–50 residues as the working definition, while acknowledging that some literature extends the upper boundary to 100 residues. The 2–50 range is a usage convention rather than a formal IUPAC ruling, and it is applied consistently here and in Understanding Protein–Peptide Interactions.

Short linear motif (SLiM) is a contiguous sequence of 3–10 residues within a largely disordered protein region that encodes a binding signal recognised by a partner domain. SLiMs are the functional units through which protein–peptide interactions shape cellular function in a mechanistically tractable way.

Modular recognition domain is a structurally autonomous protein fold (SH2, SH3, PDZ, bromodomain, WD40, and related families) that binds a defined SLiM or PTM-decorated peptide in a groove whose geometry is conserved across paralogues.

Post-translational modification (PTM) is a covalent chemical change to a residue after translation (phosphorylation, acetylation, ubiquitylation, and methylation being the most therapeutically relevant) that creates or destroys a SLiM recognised by a downstream domain.

Protein–peptide interaction is the non-covalent association between a folded recognition domain and a peptide ligand, typically with dissociation constants in the micromolar-to-nanomolar range and a buried surface area of 600–1,200 Å².

Peptidomimetic is a synthetic molecule that reproduces the pharmacophoric geometry of a peptide interface while replacing the amide backbone with a scaffold that resists proteolysis and improves cell permeability.

Structural genomics refers to high-throughput programmes (including the Joint Centre for Structural Genomics and successor initiatives) that systematically determine protein structures at scale to populate the PDB with domain–SLiM complexes. The atomic-resolution outputs from those pipelines are catalogued in the JCSG Structure Gallery.

Key Takeaways

The same structural principle, a SLiM of 3–10 residues docking into a modular recognition domain, operates across signalling cascades, organelle trafficking, chromatin regulation, and immune recognition.
Reversibility and modularity of peptide–protein interactions make them biologically flexible and therapeutically tractable; dissociation constants in the micromolar range allow competitive inhibition without antibody-scale binding energy.
Disrupting a single SLiM–domain interface can propagate effects through entire pathways, which is why peptide interfaces have become a primary focus in drug discovery.
Structural genomics programmes and computational tools (AlphaFold-Multimer, ColabFold 2.0) have populated the PDB with hundreds of domain–peptide complexes, enabling rational design of peptidomimetics and peptide-interface inhibitors.
As of 2025–2026, no BET bromodomain inhibitor or TCR-T therapy has received EMA or MHRA marketing authorisation; the field remains in clinical-stage development despite mechanistic maturity.

References

Key studies and sources referenced in this article:

London et al. (2010) — established that reversible protein–peptide interactions mediate signal transduction and protein trafficking, and catalogued the recurring binding strategies seen across peptide–protein complexes.
Johansson-Åkhe et al. (2019) — mapped disease-associated protein–peptide interaction sites, showed that pathogenic mutations cluster at the sparse anchor residues, and set a statistical-docking baseline for proteome-wide SLiM–domain prediction.
Forbes, StatPearls (NCBI, 2023) — the operational 2–50-residue working definition of a peptide used throughout this article.
ACS Chemical Biology (2015) — showed that α/β-peptide foldamers can engage intracellular protein–protein interfaces potently enough to perturb signalling in cell culture.
Gupta et al. (2022) — pharmacophore-based design of protein fragments and constrained peptides that reproduce binding affinity while resisting proteolytic degradation.