Viva Biotech's technical leadership outlined a peptide development stack combining hybrid solid and liquid phase synthesis, ADME/PK modeling and AI-guided design. The company reports GLP-1 analogues made at kilogram scale with purity of 99% or higher, and says the chemistry can reach hundreds of…
Viva Biotech's technical leadership laid out on September 14, 2026 how hybrid synthesis, ADME/PK modeling and AI-guided design are widening the reach of peptide therapeutics, even as manufacturing bottlenecks and poor oral bioavailability constrain the field. Dr. Susan Chen, the company's chief technology officer, and Dr. Yue Qian, vice president and head of Viva's Multi-Modality AI-Rooted Solutions MARS division, described a development stack that runs from solid phase peptide synthesis through computational impurity prediction to a workflow the company calls Pep2MARS . The technical case was released in an analysis written in partnership with Viva Biotech by Jennifer C. Smith-Parker, Director of Insights at BioSpace.
The central manufacturing claim is a hybrid approach that combines solid phase and liquid phase chemistry. Applied to GLP-1 analogues , the method produced material at kilogram scale with purity of 99% or higher, and the company says the same chemistry can be scaled to hundreds of kilograms. That matters because GLP-1 receptor agonists are the commercial engine behind current peptide demand, and because the standard synthesis route loses efficiency as chains grow longer.
Against that process gain sits a pharmacology problem the company does not claim to have solved. Currently marketed cyclic peptides have bioavailability below 1%, compared with 60-90% for small molecules. That gap, compounded by limited pharmacokinetic data, is the main reason oral peptide programs stall before they reach patients. The two problems are technically separate: one is a chemistry and engineering problem, the other is a pharmacology and formulation problem. Progress on the first does not imply progress on the second, and the material as presented offers no clinical readout that would connect them.
Peptide synthesis has run on Solid Phase Peptide Synthesis SPPS since the 1950s. The method builds a chain while it is anchored to a resin bead, coupling one amino acid at a time and removing a protecting group after each coupling so the next residue can attach. Between couplings the resin is washed, and side chain protecting groups are carried through the entire assembly so that only the intended amine or carboxylate is reactive at each step. The anchor is what makes the process practical: excess reagents can be flushed away rather than purified out, and the growing chain cannot be lost during workup. It is efficient for short sequences and remains the workhorse of the industry.
Length is where it breaks down. A 30-amino-acid chain requires more than 30 steps, and the final product from SPPS typically lands at 30-60% purity. Each cycle carries a yield penalty, and incomplete couplings leave deletion sequences that differ from the target by a single residue, which are difficult to separate because they behave almost identically to the desired peptide. The failures compound in ways that are not simply additive. As the chain grows, it can fold back on the resin and aggregate through backbone hydrogen bonding, burying reactive sites and slowing coupling at exactly the residues that are hardest to reach. Sequences prone to beta sheet formation are the worst offenders. The result is a crude mixture carrying truncated sequences, deletion sequences, epimerized residues that racemized during activation, oxidized side chains, and aspartimide byproducts, all present at concentrations that strain the resolving power of preparative chromatography.
Scale adds cost on top of yield loss. Large quantity production with SPPS requires reactors of 2,000-5,000 liters, and building that capacity carries a price tag in the tens of millions of dollars. Those vessels are also unforgiving of process variation. Temperature is a particular vulnerability: within a 5°C temperature range, impurities can increase threefold. Coupling kinetics, racemization rates and aspartimide formation all shift with temperature, and in a vessel of that size, heat transfer and mixing are uneven enough that different zones of the reaction can sit at different effective temperatures at the same moment. Reactor geometry feeds into the same problem, because impeller design and vessel aspect ratio determine how quickly reagents disperse and how much local excess develops during addition.
"If anything goes wrong, the money will go into the drain," Dr. Susan Chen said, describing the cost of a failed batch at that scale.
Viva's answer is to stop treating solid phase and liquid phase synthesis as alternatives. Hybrid strategies split a long target into fragments, assemble and purify those fragments where each chemistry is strongest, and then join and finish the molecule in solution, where chromatography and crystallization are more tractable than on a resin. Continuous flow processing is part of the same design philosophy, because it tightens control over mixing, residence time and temperature in a way that a large stirred vessel cannot match as precisely. In flow, the reaction volume is small and the thermal path is short, so the temperature setpoint is closer to the temperature the chemistry actually sees, and the process scales by running more reactors in parallel rather than by making one vessel larger.
The company also points to tag technology and crystallization after tag removal as tools for purification. Attaching a tag can change the physical behavior of a peptide enough to allow selective isolation, and crystallizing the tagged species can deliver a purity jump that preparative chromatography alone struggles to reach at the hundreds-of-kilograms scale, because a crystal lattice excludes impurities that would otherwise co-elute with the product. The payoff Viva reports is 99% purity or higher at kilogram-scale preparation of GLP-1 analogues, with a stated path to hundreds of kilograms. Chen's process team averages 15-20 years of peptide experience, which is relevant context for judging whether those numbers reflect a repeatable platform or a single well-run campaign, though experience is a credential rather than evidence.
None of this eliminates the underlying chemistry problem. Hybrid routes trade one set of difficulties for another: fragment condensation introduces its own epimerization risk at the junction, and the additional handling steps must be developed and validated for each sequence. What the approach offers is a different position on the yield-versus-scale curve, not a route that ignores the curve entirely.
Fragment condensation, the step that defines a hybrid route, is where most of the transferred risk lands. Joining two protected peptide segments requires activating the carboxyl terminus of one fragment, and that activated species can lose stereochemistry at the alpha carbon before the coupling completes. Racemization at a junction produces a diastereomer that is nearly identical to the target in mass and hydrophobicity, which means the impurity is often invisible to the assays used for in-process control and difficult to clear in purification. The problem gets worse with hindered residues, and it is sequence specific in a way that resists a general solution.
The fragments themselves impose constraints. Each segment must be soluble enough in the solvent system chosen for the condensation, and solubility depends on the protecting group scheme, which in turn constrains which chemistries are compatible. Segments that aggregate in solution couple poorly, so a route that solves a solid phase aggregation problem can simply reproduce it in a different physical setting. Each fragment also has to be purified, characterized and released before the next step, which multiplies the number of intermediates under control and the amount of analytical work per batch.
That is the practical meaning of a hybrid platform. It is not one process that fits all peptides but a toolbox of steps that has to be configured per molecule, and the configuration is the development work. For a company running many programs, the leverage is in reusing unit operations and analytical methods across sequences rather than in a single fixed route. For a sponsor evaluating a supplier, that distinction matters, because a demonstrated kilogram-scale batch is not the same as a demonstrated ability to reproduce it on a second, different sequence.
Peptides occupy a chemical space that small molecules and antibodies cannot. They can be orally available in principle, they can block multiple signaling pathways at once rather than inhibiting a single active site, and they can be designed for binding affinities in the picomolar range, which is the level of potency that makes a drug candidate worth developing. That multi-pathway behavior is what makes peptides interesting against protein-protein interfaces, which are large, flat and poorly suited to small molecule inhibition. Cyclization adds rigidity, which protects the backbone from proteases and can improve target engagement by preorganizing the binding conformation.
The problem is delivery. Current marketed cyclic peptides sit below 1% bioavailability, while small molecules reach 60-90%. The reasons are well characterized. Peptides are large and polar, well above the molecular weight and polar surface area ranges where passive permeability is routine. They pay a desolvation penalty to cross a lipid membrane, and they form intramolecular hydrogen bonds that satisfy polar groups internally rather than presenting them to solvent, which can help or hurt depending on the scaffold. They are substrates for efflux transporters in the gut wall and for luminal and brush border proteases, and what survives absorption is subject to first-pass metabolism in the liver. Cyclization improves stability without fixing permeability, which is why oral peptide programs often show good plasma stability in vitro and disappointing exposure in vivo.
Limited pharmacokinetic data makes the situation worse. When a program has sparse human or even animal PK, teams cannot tell whether a disappointing result reflects a molecule that never gets absorbed or one that is absorbed and then cleared too quickly. The distinction determines whether the fix is medicinal chemistry, a permeation enhancer, or a formulation change, and without an intravenous arm and an oral arm in the same study, the answer is not available. Peptide bioanalysis adds its own difficulty, because many peptide drugs resemble endogenous peptides closely enough that separating dosed compound from background requires assay work that is itself a development project, and metabolite identification is complicated when the fragments produced are amino acid sequences that occur naturally in plasma.
Viva puts the savings from AI and computational tools at 20-40% of costs, achieved by preventing wrong reactions and predicting impurities before they are made. In a process where a failed batch at large scale is expensive enough to warrant Chen's warning, avoiding a single doomed campaign changes the economics of a program. The saving is a process-economics figure, though, not a clinical one: it describes reactions not run and purification problems not encountered, and it says nothing about whether the resulting molecule works in patients.
The Pep2MARS workflow is the company's integration layer for that work, built to combine different data types to predict human pharmacokinetic parameters. Dr. Yue Qian leads the MARS division responsible for it. If the predictions hold, teams could rank candidates on projected human exposure before committing to synthesis at scale, which is the point where cyclic peptide programs…
Peptides referenced: GLP-1.
Related reading: Samsung Biologics raises $2.18B in rights offering for PolyPeptide deal, Piramal Pharma adds spray drying suite at Turbhe peptide site, Neoantigen peptide vaccines face a purity-testing standards gap, Manufacturing bottlenecks stall personalized neoantigen peptide vaccines.