Designed to Perform: De Novo Enzymes for the Next Generation of Nutrition Proteins

The next generation of medical and sports nutrition proteins will be defined less by the raw material than by the enzyme that unlocks it. Today's hydrolysates rely on enzymes borrowed from nature and tuned for generic cutting, so they still trade yield against taste and digestibility against function.

De novo enzyme design removes that trade-off. At Arzeda we design enzymes computationally, starting from the cut we want and the process it must run in. That lets us target four outcomes at once:

• Performance: peptides built for fast absorption, leucine delivery and tolerance.
• Palatability: hydrolysis that avoids the hydrophobic peptides that make hydrolysates bitter.
• Solubility and quality: hydrolysis that improves solubility at beverage pH and raises digestibility-corrected protein quality (PDCAAS).
• Feedstock freedom: enzymes that work on crude side streams — acid whey, spent grain, oilseed press cake — turning low-value waste into premium protein.

The result is protein that is better, cheaper and more sustainable at once, with malt, barley, corn, fava, wheat and oilseed proteins joining whey and soy as clinically useful ingredients.

The opportunity: hydrolysates are where nutrition science meets processing

Protein hydrolysates are a growing, premium category anchored by sports and medical nutrition, estimated at USD 3.5 billion in 2024 and USD 5.7 billion by 2034.

The clinical case rests on speed and leucine: hydrolysates are absorbed faster than intact protein in older adults, and peak plasma leucine tracks closely with muscle protein synthesis.

Whey sets the benchmark. Plant and upcycled proteins can match it — in medical supplements, where palatability drives compliance, and in sports products, where taste is the binding constraint — only if hydrolysis does far more than generic proteases can.

Why borrowed enzymes hit a ceiling

Most commercial hydrolysates use a small toolkit of natural proteases selected for availability and cost, not for the peptide profile a clinician or athlete needs. Five limits follow — for each, what happens with a generic protease, and what it costs the product.

• Bitterness. Extensive hydrolysis releases hydrophobic, proline-rich bitter peptides. Carbon or resin debittering then loses yield and essential amino acids.
• Degree-of-hydrolysis control. Past a point, smaller peptides lose emulsifying and foaming power. Digestibility is traded against texture.
• Resistant structures. Cereal prolamins — gluten, hordein, zein — resist proteases and digestion. Grain-protein yields stay low and allergenic fragments remain.
• Process mismatch. Narrow pH and temperature optima mean acid, alkaline and hot streams need costly conditioning before the enzyme will work.
• Matrix interference. Fiber, polyphenols, salts and oil inhibit enzymes, forcing higher doses and inconsistent batches.

The usual fix is to stack more natural enzymes — for example, a proline-specific endoprotease that markedly cuts casein bitterness. Reviews conclude the answer is peptidases with targeted specificity matched to process conditions. That is what design delivers.

What changes when the enzyme is designed, not found

Directed evolution improves a natural enzyme. De novo design starts from the reaction: we model the active site for the cut we want, compute sequences that hold that geometry and test the best — a design space far larger than screening can reach. As our founders put it, there is no need to improve on a starting point when you design the starting point.

For nutrition proteins, four design targets can be specified up front:

• Cleavage specificity: cut to a target peptide size without liberating bitter hydrophobic termini — debittering by design.
• Peptide-level nutrition: favor rapidly absorbed di- and tripeptides and leucine-bearing fragments.
• Epitope destruction: target the proline- and glutamine-rich allergenic sequences in cereal prolamins and beta-lactoglobulin.
• Process fit: design for the stream's real pH, temperature, solids and inhibitors, instead of conditioning the stream to suit the enzyme.

We already run this model commercially in food ingredients, including with AAK on plant-based oils. Nutrition protein is the same play, with benefits measurable in clinical endpoints.

Substrate by substrate

For each stream nutrition could draw on, here is what holds it back today, and the enzyme we would design to clear it.

• Crude and acid whey. Today: ~93–95% water and hard to dry; beta-lactoglobulin resists proteolysis. The design brief: proteases active at pH 4–5 that open beta-lactoglobulin without bitterness.
• Soy. Today: beany notes, bitterness and allergen limits. The design brief: low-bitterness, high-solubility cuts that degrade the major allergens.
• Malt and barley (brewers' spent grain). Today: ~39 Mt/yr, mostly sold as feed; proline-rich hordein locked behind fiber. The design brief: a protease–carbohydrase pairing that works on wet cake, with prolyl specificity.
• Corn (DDGS, gluten meal, stillage). Today: zein is hydrophobic and lysine-poor; food-grade zein sells at $20–40/kg against $0.20–0.30/kg for DDGS. The design brief: act on wet stillage before drying, and make zein dispersible and palatable.
• Fava. Today: good lysine, but poor solubility and beany off-notes. The design brief: acid-beverage solubility without added bitterness.
• Wheat. Today: cheap, but gluten carries proline-rich celiac epitopes. The design brief: glutenases that cut the immunogenic motifs, with whey-like kinetics.
• Other plant proteins (sunflower, rapeseed, pea, rice bran). Today: press cake at ~41–44% protein, sold as feed; phenolics inhibit enzymes. The design brief: phenolic- and oil-tolerant proteases that double neutral-pH solubility.

Even off-the-shelf hydrolysis has recovered about 65% of protein from wet brewers' spent grain, and roughly doubled sunflower protein solubility. Designed enzymes raise that ceiling.

GLP-1 is making protein quality and solubility the whole ballgame

Why quality now beats quantity. GLP-1 users eat less, and up to 40% of weight lost on semaglutide-class drugs may be lean tissue. With appetite suppressed, a protein with a mediocre quality score can no longer compensate with volume — the trend is moving from quantity to quality.

Upcycled plant proteins have historically lost here: most are limited in one essential amino acid and held back by poor digestibility, which hydrolysis raises directly.

Why solubility became non-negotiable. GLP-1 slows gastric emptying, so the format that works is a small, sippable shake delivering a full protein dose — high protein at low viscosity, often at pH 3–4. Poorly soluble protein means chalk, sediment and abandoned products.

Enzymatic hydrolysis is the established route to solubility: pepsin hydrolysis has lifted fava bean protein solubility from 24% to 89% at neutral pH. But hydrolyze far enough for acid-stable solubility and you release bitter peptides, then lose amino acids stripping them out — lowering the very quality score you were protecting.

That trade-off is a specificity problem, and specificity is designable: one enzyme can be specified for solubility at beverage pH, retained quality and gentle tolerability at once.

This is the sharpest near-term commercial case: GLP-1 companion nutrition, led by protein beverages, is forecast to reach roughly USD 16.6 billion by 2036. It needs protein that is high quality, soluble, gentle and affordable at scale — and only the enzyme decides whether upcycled streams qualify.

Where this goes next

The raw material is not the constraint. Breweries, distilleries, grain processors and oilseed crushers already generate protein at the volume nutrition needs and at a cost no purpose-grown crop can match, yet sell it as feed. A designed enzyme is a low-capital way to move those streams to nutrition-grade value — without new land, crops or water.

The question is no longer whether the protein is good enough. It is which cut you want made. We welcome that conversation with partners across dairy, brewing, distilling, grain processing and oilseed crushing: you bring the stream and the target product profile; we design the enzyme.

Contact us

Sources

1. Protein hydrolysates market size and segmentation — Global Market Insights
2. GLP-1 companion foods market outlook — Meticulous Research
3. GLP-1s and the nutrition challenge — Food & Drink Technology
4. GLP-1: redesigning the future of food and nutrition — Lux Research, May 2026
5. Pennings et al., whey vs casein in older men — Am J Clin Nutr, 2011
6. Koopman et al., hydrolysate digestion and absorption kinetics — 2009
7. Edens et al., prolyl endoprotease debittering — J Agric Food Chem, 2005
8. FitzGerald & O'Cuinn, enzymatic debittering of hydrolysates — 2006
9. Belardi et al., sequential extraction from brewers' spent grain — 2025
10. Yu et al., ultrasound-assisted enzymatic extraction from BSG
11. Protein recovery from ethanol by-products and zein — Preprints, 2025
12. Karangwa et al., sunflower protein hydrolysates — 2021
13. Jakab et al., sunflower press cake composition
14. Samaei et al., faba bean hydrolysates and fortified beverages
15. Eckert et al., enzymatic hydrolysis and faba bean protein solubility — 2019
16. Gluten-degrading enzymes for celiac disease — Nutrients, 2020
17. Acid whey disposal and volume problem — UW–Madison Center for Dairy Research
18. Arzeda enzyme design platform overview
19. Arzeda designer enzymes for food and nutrition ingredients
20. Arzeda–AAK joint development agreement