How Noncanonical Amino Acids Improve Peptide Design Beyond Stability

When an active peptide has insufficient serum stability, designers often think of introducing noncanonical amino acids (NCAAs). However, if NCAAs are understood only as “amino acids that are more resistant to degradation,” their more important value will be missed, and substitutions that lack a mechanistic basis can easily result.

NCAAs can alter backbone geometry, local torsion angles, hydrogen-bonding capability, side-chain bulk, charge distribution, and cyclization mode. A single substitution may simultaneously affect the binding interface, conformational ensemble, protease recognition, membrane interactions, SPPS coupling, and purification behavior. Therefore, NCAAs are structural and chemical design variables, not merely a few additional characters that can be freely substituted into the natural amino acid alphabet. This article discusses how to select and validate these variables around an existing parent peptide; for a foundational introduction to their types and applications in drug development, see “The Role of Noncanonical Amino Acids in Drug Development”.

Define the Problem First, Then Select the NCAA

“Which NCAA should be added” is not the starting point of design. A more effective starting point is to clarify the current limitation: Is it a mapped protease cleavage site, insufficient binding affinity, excessive conformational flexibility, lack of a suitable handle for cyclization, or difficulty balancing solubility, permeability, and selectivity? Different problems correspond to different design hypotheses.

Next, critical residues should be distinguished from modifiable positions. Existing complex structures, alanine scans, conserved motifs, SAR, and activity data can be used to annotate binding hotspots and essential backbone hydrogen bonds; when no structure is available, local tolerance can also be established through a limited set of single-point substitutions. Critical positions are not absolutely unmodifiable, but they require stronger evidence and appropriate controls. Conducting this type of controlled, incremental exploration around an existing peptide is also a core principle for preserving sequence lineage in AI optimization of existing peptides.

Six Common Design Tools Are Not Interchangeable

NCAA Type	Design Objectives It May Support	Risks That Need to Be Evaluated in Parallel
D-amino acids	Disrupt protease recognition; alter local conformation and side-chain orientation	May disrupt secondary structure or binding geometry; effects are site-dependent
N-methyl amino acids	Remove a backbone NH donor; tune conformation, protease sensitivity, and permeability in certain systems	Coupling is more difficult; may generate cis–trans isomerism and lose key backbone hydrogen bonds
Aib	Favors restricted backbone conformations; can be used to tune helical or turn propensity	Relatively sterically hindered; does not improve stability or activity in all sequences
Nal	Increases aromatic surface area and hydrophobic bulk; used to explore hydrophobic pockets or π interactions	May reduce solubility and increase aggregation, nonspecific binding, and purification difficulty
Orn / Dab	Provide side-chain amines shorter than Lys; used to adjust charge position, linkage distance, or lactam cyclization geometry	Shortening the side chain may alter binding; orthogonal protection and route compatibility are required
Other bifunctional NCAAs	Introduce selectively reactive cyclization or conjugation handles	Building blocks, protecting groups, reaction selectivity, and analytical methods must be confirmed in advance

This table provides design directions, not guarantees of properties. The same NCAA may produce entirely different results depending on its position in a sequence, or when combined with neighboring residues, ring size, and the target environment.

D-Amino Acids: Stability Gains and Binding Geometry Must Be Considered Together

An L→D chirality inversion may reduce recognition of a local sequence by certain proteases, making it suitable for building a small scan near known or suspected cleavage sites. However, the substitution also changes the backbone orientation and the projection of the side chain in three-dimensional space. If the residue directly participates in binding, stabilizes an α-helix, or maintains a turn, a D-substitution may reduce affinity or rearrange the entire local conformation.

A more conservative first-round design retains the parent control, preferentially selects positions that are structurally permissible and not clear hotspots, and compares single-point substitutions rather than stacking multiple D-AAs at once. Experimentally, activity and proteolytic stability should be measured simultaneously; observing only an extended half-life does not demonstrate that the candidate is better overall.

N-Methylation: It Alters the Backbone, Not Just Adds a Methyl Group

N-methylation removes one backbone hydrogen-bond donor and affects amide bond conformation, local flexibility, and protease recognition. In some macrocycles or specific conformational systems, it may help tune membrane permeability, but this does not mean that any linear peptide becomes more membrane-permeable after N-methylation, and it certainly does not justify inferring oral availability.

Site selection should avoid positions where the NH is required to form a key hydrogen bond, and the possible cis–trans isomerism and conformational rearrangement after methylation should be examined. Chemically, N-methyl residues often reduce the efficiency of subsequent coupling; multiple adjacent sites further increase the risk of deletion peptides and complex crude products. For the relevant mechanisms and development background, see “The Important Role of N-Methyl Amino Acids in Peptide Drugs”.

Aib and Nal: One Favors Backbone Control, the Other Interface Exploration

Aib (α-aminoisobutyric acid) has an α,α-disubstituted structure that can restrict local backbone conformation and affect helical or turn propensity in appropriate sequences. It is suitable for testing hypotheses such as whether preorganizing a particular segmental conformation is beneficial, rather than serving as a universal stabilizer. If target binding requires an extended or dynamic conformation, excessive restriction may instead be unfavorable; the steric hindrance of Aib can also affect SPPS coupling.

Nal (naphthylalanine) provides a larger aromatic side chain. If the binding pocket tolerates additional bulk, it can be used to explore hydrophobic filling, π–π, or cation–π interactions; if the space is narrow, it may generate a clash. Even if affinity improves, the increased hydrophobicity introduced by Nal may reduce aqueous solubility, promote aggregation or nonspecific binding, and significantly change RP-HPLC retention and purification conditions. Therefore, Nal candidates should be compared with smaller aromatic residue controls, while solubility, crude product distribution, and safety-related indicators are monitored in parallel.

Orn and Dab: Side-Chain Length Is Itself a Geometric Parameter

Ornithine (Orn) and diaminobutyric acid (Dab) both contain side-chain amines, but their carbon chains are shorter than that of Lys. They can move a positive charge or attachment point closer to the backbone, and can be used to fine-tune the binding interface, reduce the conformational freedom of an overly long side chain, or construct lactam bridges with different spans. The differences among Orn, Dab, and Lys are not a simple matter of “the same charge means they are interchangeable”; a change of one methylene group can alter salt-bridge distance, solvent exposure, and ring strain.

In synthetic design, it must also be clearly defined which amine participates in bond formation, what orthogonal protection is used, and whether deprotection is compatible with other functional groups. Whether they are superior to Lys can only be answered by the specific sequence, geometry, and experimental results.

NCAAs Can Serve as Cyclization Handles and Can Also Fine-Tune Ring Geometry

When natural residues cannot provide suitable anchor points, NCAAs bearing orthogonal functional groups can establish side-chain–side-chain or side-chain–terminus linkages. Combining Orn or Dab with dicarboxylic acid residues of different lengths can also change the span of a lactam ring. In this context, the NCAA determines both the linkage chemistry and the structural constraint: a handle being reactive does not mean that the bioactive conformation will necessarily be retained after ring closure.

During design, anchor-point distance, side-chain orientation, ring size, linker length, target clash, and ring strain need to be compared, while neighboring topologies and linear controls are retained. A Design Guide from Linear Peptides to Cyclic Peptides further discusses when cyclization is worthwhile, as well as the differences among disulfide, head-to-tail, and lactam routes.

Binding Interfaces and Permeability Both Require Structural Context

At the binding interface, NCAAs may fill cavities not covered by natural side chains and establish new aromatic or electrostatic interactions, but they may also create clashes due to changes in size or charge. Structural models can be used to exclude clearly unreasonable designs and compare side-chain orientation, but scoring errors, receptor flexibility, and water-mediated effects mean that models cannot guarantee improved affinity.

Permeability is likewise not determined by a single residue. Net charge, exposed polarity, intramolecular hydrogen bonds, hydrophobic surface area, conformational dynamics, and molecular size collectively affect transmembrane behavior. N-methylation, cyclization, or hydrophobic NCAAs may improve the profile in specific combinations, or they may sacrifice solubility and increase membrane toxicity. Therefore, permeability must be interpreted together with solubility, activity, and safety experiments.

From Design to Synthesis: Being Coupled into the Sequence Does Not Mean a Purified Product Can Be Delivered

NCAA designs must undergo review within a real SPPS route. First, it is necessary to confirm whether the protected building block is available, whether the protecting groups are compatible, the risks of racemization or side reactions, and the required coupling conditions; next, steric hindrance, on-resin aggregation, consecutive hydrophobic residues, and multiple modifications should be evaluated for their potential to cause deletion peptides. Cyclization projects must also consider the quality of the linear precursor, dilution conditions, intermolecular reactions, and isomer complexity.

The final question is not only “whether the target ion peak can be formed,” but also the target proportion in the crude product, whether byproducts and isomers can be separated, whether preparative HPLC has a reasonable window, and whether sufficient isolated yield can be obtained. Nal may cause overly strong retention, multiple N-methyl sites may broaden the crude product distribution, and complex cyclization may generate closely related isomers. These risks should be included in scoring during the ranking stage, rather than being handed over to the synthesis team after the sequence has been finalized. The synthesizability screening topic describes how SPPS, purification, and yield risks can be moved upstream into AI candidate selection.

NCAA-Aware Modeling Cannot Be Just “20 Amino Acids Plus Extra Tokens”

A model suitable for NCAA-containing peptides needs to represent chemical and structural information beyond residue identity, including chirality, backbone substitution, functional groups, charge, steric volume, local torsional preferences, hydrogen-bonding capability, and modification context. When sufficient structural information is available, the effects of NCAAs on local geometry, the binding interface, and the conformational ensemble should also be evaluated. At the same time, candidates must be connected to practically available building blocks and permitted chemical routes.

Therefore, experiment-oriented design is better suited to starting from a validated, synthesis-ready set of NCAAs. This allows reagent availability, protection strategy, analytical expectations, cost, and lead time to be understood in advance, rather than using open-ended generation of structures that are theoretically novel but cannot be purchased or lack a reliable route. AI can help generate controlled substitutions, compare structural hypotheses, and rank risks, but it cannot reliably predict all NCAA effects, nor can it replace synthesis and experimentation.

Multi-Objective Ranking Is Preferable to Searching for the “Highest NCAA Score”

NCAA optimization usually involves conflicting objectives. Nal may improve certain hydrophobic interfaces while reducing solubility and increasing purification difficulty; N-methylation may alter the stability or permeability profile while increasing synthetic difficulty and changing affinity; D-AAs may improve degradation resistance near a particular cleavage site while perturbing binding geometry. Compressing these results into a single overall score can hide important trade-offs.

A more reasonable approach is to retain a set of differentiated candidates: some biased toward stability, some prioritizing retention of binding, and some more robust in terms of synthetic feasibility. The experimental set should cover different mechanistic hypotheses, rather than only synthesizing several highly similar “top-scoring” sequences.

Recommended NCAA Design–Experiment Workflow

An executable workflow can be summarized as follows:

Existing peptide → Define optimization objectives → Annotate critical and modifiable positions → Select appropriate NCAA categories → Generate controlled substitutions → Sequence and structural evaluation → Synthesizability review → Multi-objective ranking → Synthesize diverse candidates → QC → Experimental testing → Next round of optimization

NCAAs should be introduced selectively and iteratively. In the first round, use single-point substitutions or a small number of combinations as much as possible to answer clearly defined questions, and use the parent peptide and necessary natural amino acid variants as controls. After HPLC and LC-MS confirm identity and purity, feed activity, stability, solubility, permeability, or other project-relevant data back into the next round. Structural models and AI narrow the experimental space; they do not replace experimental conclusions.

Conclusion

The value of noncanonical amino acids extends far beyond improving protease stability. D-AAs, N-methyl amino acids, Aib, Nal, Orn, Dab, and bifunctional NCAAs each provide different dimensions of control, including chirality, backbone hydrogen bonding, conformational preference, aromatic interfaces, side-chain geometry, and conjugation chemistry. Truly effective optimization means that each substitution should correspond to a clear hypothesis while also being evaluated for binding, stability, permeability, solubility, synthesis, and purification.

Apollomics can perform D-AA substitution, N-methylation, Aib and aromatic NCAA design, Orn/Dab side-chain engineering, cyclization handle design, structure-aware NCAA optimization, and synthesizability review based on existing candidates, and integrate these efforts with custom peptide synthesis, purification, and quality analysis. The service workflow is Existing Peptide → NCAA Design → Structural/Chemistry Review → Synthesis → QC → Experimental Iteration, and the specific plan is determined by project objectives, available evidence, and experimental feedback.