Cyclic peptides open new drug frontiers, if manufacturing can keep up

Peptide therapeutics are an increasingly hot area for drug development, fueled by the explosive GLP-1 obesity market. But peptides can pose scientific and operational challenges, such as low bioavailability and that their production is notoriously sensitive to many factors. These include reactor geometry, resin swelling and even a 5°C operation temperature shift.

In this interview with BioSpace, Viva Biotech’s Dr. Susan Chen, chief technology officer and Dr. Yue Qian, vice president and head of the Multi-Modality AI-Rooted Solutions (MARS) division, spoke about the bottlenecks in moving a peptide program from discovery into development and manufacturing.

They also laid out how AI and computational tools are changing peptide discovery and optimization in practical terms. Data is most valuable for improving predictive models in peptide discovery and process development, they said. Recent advances in cyclic peptide design and modeling change the opportunity for peptide therapeutics.

Peptides are small chains of amino acids linked by chemical bonds. Complex or long-chain peptides are particularly difficult to develop and produce reproducibly due to obscure reaction mechanism, unpredictable impurities during the scale‑up process and difficult structure characterization.

Peptides are very fragile: sensitive to reactor geometry, resin bed height, mixing efficiency, stirring speed and temperature. Within just a 5°C range, impurities can increase threefold, making reproducibility paramount for commercial products, Chen explained. Reproducibility, process and impurity control are extremely important for commercial products.

Peptide synthesis began in the 1950s, using a technique known as Solid Phase Peptide Synthesis (SPPS), Chen said.

SPPS is a heterogeneous process involving a resin base and liquid-phase reagents. As an amino acid chain becomes longer, coupling efficiency drops and side product formation increases. This creates significant purification challenges. For a 30‑amino‑acid chain, for example, SPPS requires more than 30 steps, with a low yield result and the purity of the final product hovering around a mere 30–60%. Therefore, big SPPS reactors, such as 2,000~5,000 liters, are required for large quantity of productions, costing tens of millions of dollars.

Column purification, followed by lyophilization, a method used to separate and isolate a single target chemical, is commonly used to obtain final peptide products. However, it is costly and time‑consuming, and peptides stored in the liquid phase degrade slowly when the purification process takes a longer time.

“If anything goes wrong, the money will go into the drain,” Chen said.

To mitigate these prohibitive risks, Viva Biotech’s peptides, proteins and oligonucelotides team has developed and applied hybrid synthesis strategies for peptide development, integrating solid-phase and liquid-phase approaches to support more efficient synthesis and scale-up.

The platform has applied this hybrid approach to GLP-1 analogues, with kilogram-scale preparation at purity of 99% or higher. It has also used combined SPPS/liquid phase peptide synthesis strategies for complex peptide programs.

Hybrid methods combine solid and liquid phases, producing fragments with high purity and bigger throughput, which can be scaled to hundreds of kilograms. The final peptide can be produced by the coupling of these fragments. This hybrid approach reduces costs and minimizes risk in production.

It’s crucial for the discovery stage of peptide development to be in lockstep with manufacturing, even though the latter is often seen as an afterthought to drug likeness or potency, Qian said. Failures at small scale lead to greater issues in later stages of the manufacturing process, Qian and Chen agreed.

Quality by Design and a Failure Mode and Effects Analysis model, a step-by-step risk management tool, are essential in process understanding, Chen said.

Qian’s modeling team works closely with Chen’s group to design peptides that can be easily synthesized with less chiral centers and more effective reactions. They also collaborate on breaking rings of cyclic peptides and assembling smaller pieces, noting that even the same peptide can have different synthesized routes.

In the complexity of moving from peptide discovery to development, bioavailability for peptides is a sticking point, and especially for oral peptides, Qian said. Cyclic peptides can be orally available like small molecules, but current marketed options have bioavailability of less than 1%, compared to 60–90% for small molecules.

A compounding factor for cyclic peptides is very limited data points on pharmacokinetic (PK) properties, Qian said. Viva Biotech considers data points in literature and patents, as well as in mice and rat data, to transfer learning algorithms for human predictions. This allows for better bioavailability predictability and cost-savings as less money can be spent on producing active pharmaceutical ingredients, she added.

While high picomolar binding affinity is desirable, robust PK and clearance data are ultimately more critical, Qian said. Peptides degrade quickly due to enzymes like DPP‑4, so in vitro stability assays and drug absorption, distribution, metabolism and elimination (ADME)/PK data are essential. Viva Biotech’s Pep2MARS workflow covers epitope prediction, de novo design, optimization, permeability, stability, toxicity and bioavailability.

Chen added that critical process parameter is also critical for building predictive models.

AI and computational tools can save “huge” amounts of money, in the scale of 20-40%, in preventing wrong reactions and predicting impurities, Chen said. A single incorrect amino acid sequence can compromise the entire batch, highlighting the need for better understanding of whole process and high risks during production.

AI also enables multiparameter optimization, turning sequential discovery into a process where all properties are evaluated in parallel. These models can sample conformations and predict ADME/PK profiles.

In terms of automated processes, Qian said solid‑state synthesis can be automated but some cyclic peptides require liquid-phase synthesis. Overall, the peptide space still requires extensive human interaction and practical training, Chen said, noting her team averages 15–20 years of peptide experience.

As for cyclic peptide opportunities, Qian explained they provide a unique advantage relative to small molecules and antibodies. Many targets are not druggable by small molecules due to flat surfaces, and antibodies cannot cross cell membranes or be orally available. Cyclic peptides can block multiple signaling pathways, provide selectivity and even replace antibody therapies with oral administration.

Viva Biotech uses DNA-encoded libraries, phage display, mRNA display and de novo AI design to generate hits and improve models based on wet‑lab feedback. These approaches have seen great success with high hit rate and novel hits advancing into the later stages.

In the future, the landscape should expect more investment in high‑throughput automation and more building blocks and amino acids entering the market, Qian said. Viva Biotech’s development of novel synthesis technologies, including continuous flow, TAG technology, crystallization after tag and new modalities like PDC, APDC, RDC and POA will also enhance the field, Chen added.

By combining AI-enabled peptide discovery and optimization with capabilities in peptide synthesis, process development, analytics, formulation and manufacturing, Viva Biotech’s CRDMO platform is designed to support programs from early discovery through scalable development and commercial manufacturing.

This article was written in partnership with Viva Biotech.