Peptide-to-Small-Molecule Translation

Peptides occupy a middle ground in modern drug discovery: they can access protein-protein interfaces and dynamic surfaces that small molecules typically cannot, while remaining more tunable than antibodies or other large biologics. Yet most peptide leads never reach development. Even highly potent binders often fail on permeability, metabolic stability or manufacturability long before the clinic (Figure 1).^1-3

Small molecules, by contrast, remain the workhorse of pharmacology – chemically tractable, stable and with many oral precedents – but they rarely engage the extended, shallow regions that peptides readily recognize. Bridging these two chemical spaces through peptide-to-small-molecule (pep-to-SM) translation aims to combine the biological insight of peptides with the developability of small molecules. What began as a creative idea has become an increasingly systematic design strategy.⁴

Why the Approach Makes Sense

When a peptide binds a target, it provides a high-resolution map of how recognition occurs: charged pairs, hydrogen-bond networks, aromatic contacts and hydrophobic anchors. If those interaction roles can be recast into a smaller scaffold while preserving their spatial and chemical logic, teams can move from validated biology to viable chemistry.^5,6

Several systems illustrate this principle in practice. In the p53-MDM2 axis, the co-crystal structure of the p53 transactivation-domain peptide with MDM2 revealed a compact hydrophobic pocket that accommodates a triad of p53 residues (Phe19, Trp23, Leu26) (Figure 2). This hot-spot pattern directly inspired small-molecule antagonists such as the Nutlin series, designed to occupy the same pocket and mimic those key side-chain interactions in a drug-like scaffold. A similar trajectory is seen for SMAC-IAP: the N-terminal AVPI motif of SMAC binds into a defined groove on IAP BIR domains, and that peptide interaction map enabled the development of small molecule “SMAC mimetics” that reproduce the AVPI contact pattern and have progressed into clinical testing. In both cases, structural biology turned a peptide binding mode into a functional map that guided successful small molecule design.^7,8

The strategy does not guarantee success, but it frequently shortens discovery cycles by grounding chemical design in experimentally proven biology.

What Pep-to-SM Translation Actually Means and Why it Often Fails

Effective translation is not a matter of simply “shrinking” a peptide. It requires identifying the functional pharmacophore – the minimal set of interaction roles essential for binding – and reconstructing those functions within a small molecule framework that satisfies drug-like constraints.

Figure 1: Comparative overview of therapeutic modalities

Source: Receptor.AI

Typical workflows involve:

• Analyzing peptide-protein structures to define critical contacts
• Mapping and ranking target subpockets capable of sustaining those interactions
• Encoding the key electrostatic and hydrophobic roles into a pharmacophore model
• Enumerating or screening small molecule scaffolds predicted to reproduce those functions

Success depends less on visual similarity to the original peptide than on reproducing its interaction logic. Structure does not equal function — the goal is to recreate chemistry, not to copy coordinates.⁵

Figure 2: Peptide-guided mapping of the p53-MDM2 pocket. The p53 transactivation-domain peptide (red helix/green sticks) binds MDM2 via a hydrophobic triad (Phe19, Trp23, Leu26) buried in a well-defined cleft. This hot-spot pattern forms the structural basis for later small molecule MDM2 inhibitors that mimic these key interactions.

Source: Zhu, et. al. 1 (2022). Targeting p53–MDM2 interaction by small-molecule inhibitors: learning from MDM2 inhibitors in clinical trials. J Hematol Oncol 15, 91.

Three Common Failure Modes

1. Geometry ≠ function

Maintaining a similar 3D shape doesn’t ensure equivalent binding. True activity depends on preserving the electrostatic and hydrogen-bonding roles across realistic conformational states, not on static atom-by-atom overlays.

2. No anchorable subpockets

Many peptide targets – particularly PPIs – lack well-defined cavities. Without anchor points, small molecules drift toward excessive polarity (hurting ADME) or excessive hydrophobicity (hurting selectivity). The first design question should always be: is there an anchor point?

3. Single-structure bias

Most designs still rely on one crystal or cryo-EM snapshot. Real interfaces are dynamic, with side-chain rearrangements and cryptic pockets that appear only under certain conditions. Optimizing to one frame of a flexible system usually leads to collapse during validation.^9,10

Each of these issues stems from the same misconception: treating peptide recognition as static rather than dynamic.

Modern Fixes — Functional Translation Under Dynamic Constraints

Recent advances address these limitations by reframing peptide-to-SM translation as a functional mapping problem governed by molecular dynamics and property constraints, including:

• Functional pharmacophore extraction: Algorithms now capture electrostatic, hydrogen-bonding and hydrophobic interactions rather than geometric features, ensuring chemical logic is preserved even when the backbone is lost.^5,6
• Subpocket identification and ranking: Machine-learning and structure-based scoring tools help identify which surface features can actually sustain small molecule anchoring, allowing chemists to focus on realistic binding niches.¹¹
• Dynamic ensemble modeling: Instead of relying on a single rigid structure, ensemble docking across multiple conformational states accounts for protein flexibility and induced fit, significantly improving hit quality.
• AI-guided property triage: After candidate scaffolds are generated, integrated models evaluate permeability, polarity, and stability alongside binding potential, concentrating synthesis on compounds likely to exhibit genuine drug-like behavior.

Figure 3: Overview of Receptor.AI’s peptide-to-small-molecule translation workflow

Credit: Receptor.AI

An increasing number of platforms now integrate these components end-to-end. For example, Receptor.AI’s peptide-to-SM translation workflow encodes peptide pharmacophores into a latent representation, identifies and ranks subpockets with a help of dynamic ensemble modeling, and decodes the result into small-molecule hypotheses for AI-guided screening (Figure 3). The approach enables functional peptide mimetic design even for proteins with no known small molecule ligands.

Together, these advances are turning what was once a trial-and-error exercise into a disciplined and reproducible design workflow.

A Pragmatic, Not Universal, Strategy

Peptide-to-SM translation is best viewed as a target-specific strategy, not a universal solution. It succeeds when three conditions are met: (1) the peptide’s essential pharmacophore can be expressed compactly, (2) the target surface provides stable, rankable subpockets, and (3) designs are evaluated across realistic conformational ensembles.

When these conditions are met, modern computational and AI-assisted pipelines – combining pharmacophore encoding, pocket analytics, dynamics, and multi-property screening – can move teams efficiently from biological binders to chemically tractable leads.

References

1. Lamers, C. (2022). Overcoming the shortcomings of Peptide-Based therapeutics. Future Drug Discovery. 4(2).
2. Buyanova M, Pei D. Targeting intracellular protein–protein interactions with macrocyclic peptides. Trends in Pharmacological Sciences. 2021;43(3):234-248.
3. Sharma, K. et. al. (2022). Peptide-based drug discovery: Current status and recent advances. Drug Discovery Today. 28(2):103464.
4. Wu, D., et. al. (2023). Small molecules targeting protein-protein interactions for cancer therapy. Acta Pharmaceutica Sinica B. 13(10):4060-4088.
5. Han, Z., et. al. (2024). Transformation of peptides to small molecules in medicinal chemistry: Challenges and opportunities. Acta Pharmaceutica Sinica B. 14(10):4243-4265.
6. Hayward, D. et. al. (2024). Strategies for converting turn-motif and cyclic peptides to small molecules for targeting protein–protein interactions. RSC Chemical Biology. 5(3):198-208.
7. Zhu, H. et. al. (2022). Targeting p53–MDM2 interaction by small-molecule inhibitors: learning from MDM2 inhibitors in clinical trials. Journal of Hematology & Oncology. 15(1):91.
8. Wang, S. (2010). Design of Small-Molecule SMac Mimetics as IAP antagonists. Current Topics in Microbiology and Immunology. 348:89-113.
9. Bemelmans, M., et. al. (2025). Computational advances in discovering cryptic pockets for drug discovery. Current Opinion in Structural Biology. 90:102975.
10. Basciu, A. et. al. (2022). No dance, no partner! A tale of receptor flexibility in docking and virtual screening. Annual Reports in Medicinal Chemistry. 43-97.
11. Alzyoud, L., et. al. (2022). Structure-based assessment and druggability classification of protein–protein interaction sites. Scientific Reports. 12(1):7975.