Cyclic vs Linear Peptides: Why Shape Matters for Function
Peptide Modifications
66 approved cyclic peptide drugs
As of 2024, 66 cyclic peptide drugs have been approved globally, with 39 gaining approval since 2000, reflecting accelerating interest in constrained peptide architectures.
Ji et al., Angewandte Chemie, 2024
Ji et al., Angewandte Chemie, 2024
View as imageA linear peptide is a floppy chain. It can adopt thousands of conformations in solution, most of which are not the one needed to bind a target. It exposes its N- and C-termini to exopeptidases, the first enzymes to attack. And its extended surface area gives endopeptidases plenty of cleavage sites. A cyclic peptide eliminates or reduces all three problems: it constrains conformation, removes vulnerable termini, and shields backbone amide bonds. These structural advantages have driven 66 cyclic peptide drugs to regulatory approval globally, with the pace of approvals accelerating since 2000. For a broader view of modification strategies that improve peptide stability, see the pillar article on peptoid modifications that resist protease degradation.
Key Takeaways
- Cyclization constrains peptide conformation, reducing the entropic cost of target binding and typically increasing affinity by eliminating non-productive conformations
- D-amino acid substitution combined with cyclization produced antimicrobial peptides that maintained activity in serum, where the linear parent peptide lost effectiveness (Mendes et al., 2026)[1]
- A cyclic peptide carrier enabled oral delivery of insulin hexamers with sustained glycemic efficacy in diabetic models, achieving what linear peptide carriers cannot (Chikamatsu et al., 2026)[2]
- AlphaFold2 has been adapted to predict and design cyclic peptide structures, with designed peptides achieving interaction scores comparable to known protein-protein interaction binders (Halbwedl et al., 2026; Rettie et al., 2025)[3][4]
- Bicyclic peptides represent the frontier, enabling targeted protein degradation of extracellular targets through CPPTAC technology (Bai et al., 2026)[5]
- Cyclic peptides still generally show oral bioavailability below 2%, though recent de novo designs have reached 18% in rats
The structural differences
Linear peptides are synthesized as chains from N-terminus to C-terminus. They are flexible, typically unstructured in solution, and present both termini as recognition sites for exopeptidases. Their conformational freedom means they sample many shapes, and the probability of being in the "right" shape for target binding at any given moment is low.
Cyclization connects the ends. The most common approaches:
- Head-to-tail (backbone) cyclization: The N-terminus bonds to the C-terminus, creating a macrocyclic ring. This eliminates both terminal recognition sites simultaneously. Wan et al. (2026) developed automated rapid synthesis of head-to-tail cyclic peptides using a diaminonicotinic acid scaffold, producing high-purity products in a streamlined process.[6]
- Side-chain to side-chain cyclization: A bridge forms between amino acid side chains (often cysteine disulfide bonds or lactam bridges between lysine and glutamic acid). The termini remain free but the internal constraint stiffens the backbone.
- Stapling: A hydrocarbon bridge connects non-adjacent residues on the same face of a helix, locking the peptide into an alpha-helical conformation. Wang et al. (2025) demonstrated that a short double-stapled peptide mimicking the HR2 core region potently inhibited human betacoronaviruses.[7] For more on this approach, see Stapled Peptides: Locking Peptides into Helical Shapes for Drug Design.
Each strategy produces different ring sizes and conformational constraints. The design choice depends on which structural feature matters most for the target interaction. For the chemistry behind these methods, see Cyclization: How Closing the Ring Stabilizes Peptides.
Protease resistance: the primary advantage
The single biggest limitation of linear peptides as drugs is proteolytic degradation. Serum half-lives of unmodified linear peptides are typically measured in minutes. Cyclization addresses this through multiple mechanisms: removing exopeptidase recognition sites (terminal amino acids), constraining the backbone so endopeptidase cleavage sites are geometrically inaccessible, and reducing conformational flexibility so the peptide cannot unfold into a protease-accessible shape.
Wang et al. (2026) provided a direct demonstration: they took a beta-turn antimicrobial peptide with potent activity against multidrug-resistant bacteria but poor serum stability, then applied both cyclization and stereochemical inversion (D-amino acid substitution). The resulting cyclic peptide maintained antimicrobial activity while overcoming the proteolytic instability that rendered the linear version clinically useless.[8]
Mendes et al. (2026) quantified this further with arginine-rich antimicrobial peptides: among eight linear peptides tested, R4F4 showed the strongest antibacterial activity, but its effectiveness was significantly reduced in the presence of serum proteases. D-amino acid substitution combined with cyclization preserved the antimicrobial function that the linear form lost under physiological conditions.[1]
The macrocyclic peptide rhesus theta defensin-1 (RTD-1) illustrates how nature uses cyclization for the same purpose. This backbone-cyclized peptide, found in rhesus macaque neutrophils, activates interferon and antiviral pathways in human monocytes. Its fully cyclized backbone gives it exceptional stability compared to the open-chain defensins found in humans.[9] The ultra-stable cyclotides found in plants represent the most extreme version of this principle.
The magnitude of the stability improvement varies by sequence and cyclization method, but the pattern is consistent. Comparative studies typically report 10- to 100-fold increases in serum half-life for cyclic versus linear versions of the same sequence. In extreme cases, particularly with cystine-knot architectures that combine head-to-tail cyclization with multiple disulfide bonds, the stability increase can exceed 1,000-fold. Cyclotides from the plant Oldenlandia affinis, for example, survive boiling water, gastric acid at pH 2, and exposure to multiple proteases simultaneously.
Binding affinity and selectivity
Cyclization improves target binding through a thermodynamic mechanism: pre-organizing the peptide into its bioactive conformation reduces the entropic penalty of binding. A linear peptide that must fold from a random coil into a binding-competent shape pays an entropic cost. A cyclic peptide that is already constrained into approximately the right shape pays less.
Rowland et al. (2025) reviewed rational design of cyclic and bicyclic peptides, concluding that these structures "offer better stability and bioavailability than linear peptides while keeping the ability to block protein-protein interactions that small molecules cannot reach."[10] This is the key positioning: cyclic peptides occupy a molecular weight and surface area sweet spot between small molecules (which lack the contact surface to disrupt protein-protein interactions) and antibodies (which are expensive and cannot enter cells).
This positioning has attracted pharmaceutical investment. The 66 approved cyclic peptide drugs span indications from immunosuppression (cyclosporine) to antibiotics (daptomycin, vancomycin, polymyxin B) to hormonal disorders (octreotide, lanreotide) to obstetrics (oxytocin, carbetocin). The diversity of indications demonstrates that the structural advantages of cyclization translate across therapeutic areas rather than being limited to specific target classes. The approval rate has nearly doubled since 2000, driven by better cyclization chemistry, computational design tools, and growing interest in protein-protein interaction targets that small molecules struggle to address.
The binding affinity advantage is not guaranteed, however. Cyclization constrains the peptide into a specific conformation, and if that conformation does not match the target-bound state, affinity can decrease. This is why systematic comparison studies, like those by de Veer et al. with different stapling strategies, are essential: the optimal cyclization for one target may be suboptimal for another. Peptide engineers must balance stability gains against potential affinity losses, and the choice of cyclization chemistry (head-to-tail, disulfide, staple position) directly determines this trade-off.
Bicyclic peptides push this further. Bai et al. (2026) developed bicyclic peptide-based CPPTACs (cell-penetrating peptide targeted chimeras) that selectively degrade extracellular and cell membrane proteins through CPP-induced endocytosis and lysosomal delivery.[5] The second ring provides additional constraint and an additional binding surface, enabling bifunctional designs that a single-ring cyclic peptide cannot achieve.
Membrane permeability: the persistent challenge
Cyclization does not automatically make peptides cell-permeable. This is a common misconception. While constraining the backbone can bury polar amide bonds through intramolecular hydrogen bonding (reducing the desolvation penalty of crossing a lipid bilayer), most cyclic peptides remain poorly membrane-permeable.
Nielsen et al. (2025) measured membrane permeability for random cyclic peptides in living cells and found that the relationship between structure and permeability was not straightforward, establishing baseline data to guide drug development.[11] Li et al. (2026) showed that cyclization of a beta-hairpin peptide, combined with thermal modulation, enabled efficient cellular delivery, but the cyclization alone was insufficient.[12]
The oral bioavailability picture is similarly nuanced. Most cyclic peptides achieve oral bioavailability below 2%. Recent de novo design efforts have pushed this to 18% in rats, a substantial advance but still well below the 50%+ threshold typical of successful oral drugs. Chikamatsu et al. (2026) demonstrated a creative workaround: using a small-intestine-permeable cyclic peptide (DNP-V) not as the drug itself but as a carrier to ferry insulin hexamers across the gut epithelium, achieving robust glycemic efficacy in diabetic models.[2] For the broader oral peptide pipeline, see The Future of Oral Peptide Drugs: What's in the Pipeline.
Computational design: AI meets cyclization
The design space for cyclic peptides is enormous. A 10-residue cyclic peptide has 20^10 possible sequences (~10 trillion), each with multiple possible ring closures, stereochemistries, and N-methylation patterns. Rational design cannot sample this space effectively. Computational methods are transforming the field.
Rettie et al. (2025) adapted AlphaFold2 to predict cyclic peptide structures, publishing the method in Nature Communications. The adapted model successfully predicted three-dimensional structures of cyclic peptides and generated de novo designs.[4] Halbwedl et al. (2026) extended this to protein-protein interaction targeting, using AlphaFold2-guided design to create cyclic peptide stabilizers that achieved interaction scores comparable to known PPI binders.[3]
These tools are converging with automated synthesis. Wan et al. (2026) developed automated head-to-tail cyclic peptide synthesis that produces high-purity products rapidly, enabling experimental validation of computationally designed candidates at scale.[6]
The combination of AI-based structure prediction, automated synthesis, and high-throughput permeability screening (as demonstrated by Nielsen et al.) creates a design-make-test cycle for cyclic peptides that did not exist five years ago. The practical impact is already visible: computational methods that once required months of supercomputer time to predict a single structure can now screen thousands of candidates in hours, prioritizing those with the highest predicted binding affinity, membrane permeability, and protease resistance for experimental validation.
This convergence also addresses a historical bottleneck in cyclic peptide development. Previous design approaches relied heavily on natural product scaffolds (cyclosporine, daptomycin) or systematic alanine scanning of known active sequences. AI-guided design can explore regions of chemical space that nature never sampled, potentially identifying cyclic peptides with properties, like oral bioavailability exceeding 10%, that natural scaffolds rarely achieve.
The trajectory from natural discovery to computational design mirrors what happened with antibody engineering decades earlier. Antibodies began as natural molecules, progressed through hybridoma selection, then phage display, and finally computational design. Cyclic peptides are following the same path, with the added advantage that their smaller size makes computational structure prediction more tractable. A cyclic decapeptide has roughly 1,000 atoms; a full antibody has over 20,000. This size advantage means that cyclic peptide design tools can achieve higher accuracy and throughput than antibody design tools, potentially compressing what took antibody engineering three decades into a single decade for cyclic peptides.
Hyun et al. (2025) highlighted cyclotides as natural scaffolds that bridge the computational and experimental worlds: their head-to-tail cyclization plus cystine knot architecture creates a rigid, protease-resistant framework onto which diverse pharmacophores can be grafted. The cyclotide scaffold tolerates extensive sequence variation in its surface-exposed loops while maintaining its structural integrity, making it a natural starting point for computational library design.[9]
When linear peptides are better
Cyclization is not universally superior. Linear peptides retain advantages in several contexts:
Manufacturing simplicity. Linear peptides are cheaper and faster to synthesize. Solid-phase peptide synthesis of linear chains is well-established at industrial scale; cyclization adds steps, reduces yields, and introduces purification challenges (separating cyclic from oligomeric byproducts).
Flexibility as a feature. Some targets require conformational adaptability. Intrinsically disordered protein regions often bind through induced fit, where the peptide adopts its binding conformation only upon contact with the target. Pre-constraining with cyclization can prevent this.
Longer sequences. Cyclization works best for peptides of 5-15 amino acids. Larger rings become increasingly flexible and lose the conformational constraint that makes cyclization beneficial. For longer peptides, other strategies like D-amino acid substitution or beta-peptide backbones may be more appropriate.
Prodrug strategies. Linear peptides can be designed as prodrugs that are activated by proteolytic cleavage at the target site. Cyclization, which prevents cleavage, would defeat this purpose.
Cost and accessibility. For research applications where hundreds of peptide variants need to be screened, the additional cost and complexity of cyclization per variant can be prohibitive. Linear peptide libraries are faster to generate and screen, making them preferred for early-stage discovery even when the final therapeutic candidate may be cyclized.
The choice between cyclic and linear is ultimately a design decision driven by the specific therapeutic context: what target, what route of administration, what duration of action, and what manufacturing constraints apply. In practice, many drug development programs start with a linear peptide lead and cyclize it later in optimization, treating cyclization as one tool in a broader modification toolkit that includes peptidomimetic approaches and unnatural amino acid substitutions. The decision is rarely binary: hybrid strategies that cyclize one region while leaving another linear are increasingly common, combining the stability benefits of cyclization with the flexibility needed for certain binding interactions.
The Bottom Line
Cyclic peptides offer substantial advantages over linear peptides in protease resistance, target binding affinity, and conformational stability, with 66 approved drugs demonstrating clinical viability. Cyclization does not automatically confer membrane permeability or oral bioavailability, though computational design tools (AlphaFold2, machine learning models) and automated synthesis are rapidly expanding the accessible design space. Linear peptides remain preferable when manufacturing simplicity, conformational flexibility, or prodrug activation is required. The frontier lies in bicyclic architectures and AI-guided design of orally bioavailable cyclic peptides.