Enzyme Class Is Table Stakes: Design for the Process, Not the Lab

For decades, enzyme discovery followed a familiar path: search the natural databases, screen broad enzyme families, find a candidate that performs the desired reaction. Landing in the right enzyme class felt like the milestone that turned a lab concept into a commercial candidate. In practice, it is only the starting line. The real test comes at scale, where high substrate concentrations, higher temperatures, pH swings, organic solvents, and intense mixing push natural enzymes outside the conditions they evolved for. Computational de novo design reverses the order: start from the manufacturing specification, and build the enzyme to meet it.

Natural enzymes evolved to work inside living organisms. Pushed outside that comfort zone, they often lose activity, wear out quickly, and stop making economic sense.

How the industry usually closes the gap

Manufacturers typically bridge the distance between a promising enzyme and a profitable process in one of two ways:

• Bend the process to suit the enzyme. Run cooler, mix more gently, or dilute the feed to keep a fragile enzyme alive. This works, but it costs throughput and often means larger tanks, more water, and more energy.
• Retrofit the enzyme after discovery. Use protein engineering, such as directed evolution, to toughen a natural enzyme for industrial conditions without damaging the part that does the chemistry.

Retrofitting can deliver major wins. The best-known example is Merck and Codexis's engineered transaminase for the diabetes drug sitagliptin. Through many rounds of laboratory evolution, the team improved the enzyme's activity more than 25,000-fold. The enzymatic route replaced a high-pressure, rhodium-catalyzed step, increased overall yield by 10–13%, raised productivity by 56%, and cut total waste by 19%, earning a 2010 Presidential Green Chemistry Challenge Award [1, 2].

But retrofitting is slow, and it is limited by the natural starting point. Both approaches share the same mindset: they treat manufacturing process conditions as an obstacle to work around rather than as a design specification.

Flipping the approach: start from the process conditions at scale

Computational de novo enzyme design reverses the order. Instead of finding an enzyme and then adapting the process or the protein, designers start with the manufacturing specification — target pH, temperature range, solvent, and operating mode — and build an enzyme from scratch to meet it.

Combining physics-based modeling with generative AI lets designers build two requirements into a single protein from the start:

• The right chemistry. An active site precisely shaped to carry out the desired reaction.
• The right toughness. An overall structure tuned for the pH, temperature, solvent, and continuous-operation demands of the process at scale.

The underlying science is advancing quickly. In 2025, University of Washington researchers reported in Science AI-designed enzymes with complex active sites unlike anything in nature; their laboratory structures matched the computer designs to within one angstrom [3]. The field is still maturing, and newly designed enzymes do not always match the activity of the best natural ones on the first pass, but the ability to design function and stability together is now real.

Why this matters to the bottom line

Enzymes designed for the manufacturing environment from day one can change the economics of biomanufacturing in several ways:

• More output per tank. Enzymes built for high substrate loads and process temperatures increase volumetric productivity, so existing assets produce more.
• Longer catalyst life. Greater thermal and chemical stability means less frequent enzyme replacement in both batch and continuous operations.
• Simpler downstream processing. Fewer protective additives and process accommodations reduce purification cost and complexity.
• Fewer scale-up surprises. When scale-up conditions are part of the design brief, there is less need to re-engineer the process or the enzyme late in development.

Three questions to ask about your current processes

• What conditions would our process ideally run at if the enzyme were not the constraint?
• How much throughput or energy are we giving up to keep our enzymes stable?
• How often do we replace enzymes, and what does that cost per year?

The bottom line

Finding the right enzyme class is table stakes. The commercial advantage goes to whoever designs the catalyst for the commercial process conditions it will actually run in — moving from model to manufacturing without the costly retrofit.

Arzeda will be at SupplySide Global in Las Vegas, October 28–30. If an enzyme is the constraint in one of your processes — or you are deciding how to make a high-value ingredient at scale — meet the team there, or start the conversation now.

• Book a meeting
• Contact us

Sources

1. U.S. Environmental Protection Agency. Presidential Green Chemistry Challenge: 2010 Greener Reaction Conditions Award (Merck & Co. and Codexis). epa.gov
2. Savile, C. K., et al. (2010). Biocatalytic asymmetric synthesis of chiral amines from ketones applied to sitagliptin manufacture. Science, 329, 305–309. doi.org/10.1126/science.1188934
3. Lauko, A., Pellock, S. J., et al., & Baker, D. (2025). Computational design of serine hydrolases. Science. doi.org/10.1126/science.adu2454