Enzyme Engineering Enters the Era of Intelligent Design: An Industrial Evolution of an Enzyme

Enzyme Engineering Enters the Era of Intelligent Design: The Evolution of an Enzyme

Enzymes are ubiquitous in our daily lives. In the soft slice of bread we eat in the morning, amylase helps make it fluffy; a bit of protease in laundry detergent allows sweat and blood stains to break down even in cold water; in brewing vats, the combination of different enzymes determines the endless variations of flavor. Even every breath we take and every thought we have is supported by thousands of enzymes operating precisely behind the scenes.

It can be said that enzymes are the most exquisite 'nanomachines' that life has bestowed upon humans. But natural enzymes are like unpolished jade—they often 'stop working' under industrial conditions such as high temperatures, strong acids or bases, or organic solvents, with insufficient catalytic efficiency, narrow substrate range, and poor stability. These limitations hinder their ability to truly change the world.

Thus, enzyme engineering emerged.

1. Methodological Leap in Enzyme Engineering

A paradigm shift in enzyme engineering

The core proposition of enzyme engineering is simple: to make a single enzyme molecule more "useful." But no matter how he changed it, he went through a long and winding road.

If you think of enzymes as a "book from the sky" written with 20 amino acid letters, early modification methods essentially did two things—either massive copying and deliberately creating typos, hoping that a single typo would actually make the sentence more beautiful; Or relying on only a half-understood grammar knowledge to alter a word they thought was key, but often ended up ruining the original meaning of the passage.

The former is directed evolution. It doesn't ask you "why," just mass-scale trial and error: using PCR and other methods to randomly scatter thousands of mutations into enzyme gene sequences, then throwing all variants into a screening system to "race." Whoever survives and runs faster wins. This method was simple yet effective, and indeed won the Nobel Prize—but at its core, it was a high-stakes gamble. The mutation space is astonishingly vast (a protein with only 300 amino acids can have nearly 6,000 single-point mutation combinations, and multi-point combinations are astronomical), while the vast majority of random mutations are either harmful or ineffective. The industry often jokingly calls this "sifting sand to find gold grains"—money-burning, time-consuming, and highly dependent on luck.

The latter is rational design. It tries to see through the structure before making decisions: it takes the enzyme's crystal structure, circles several "seemingly important" sites near the active center, and performs targeted mutations. This idea is elegant, but the problem is that proteins are not static building blocks; amino acids are separated by great distances and can be pulled at each other by charges and hydrophobicity. Changing one point often triggers a chain reaction, turning the predicted "functional gain" into uncontrollable structural disturbances. It's like thinking you've only changed one string, but the whole tone—often worsens.

The turning point came from AI. When deep learning begins to "read" billions of protein sequences, it gradually internalizes a set of "protein grammar rules" that humans find hard to express—which sites are highly conservative because they form the structural skeleton, which regions allow change because they are functional adjustment knobs, and which mutation combinations have repeatedly appeared throughout evolutionary history because they are proven good combinations. Once this syntax is mastered, the modified enzyme is no longer blind trial and error but becomes a well-founded deduction: intelligently searching the mutation space on a computer, directly identifying the most promising dozens of designs from trillions of possibilities, and then sending them for experimental verification. After the experimental data is transmitted, the model continues to learn, making the next design more accurate.

At this stage, enzyme engineering truly bids farewell to the "manual workshop" model and enters an era of computable, iterative engineering.

2. Ushering protein design into a closed-loop intelligent era

MatwingsVenus agent

At the forefront of industry practice, Shanghai Matwings Technology's independently developed MatwingsVenus™ ™ platform is a prime example of this trend. It is not a single tool but a deeply integrated enzyme molecular design system integrating protein language models, structure prediction, and closed-loop dry-wet experiments.

Simply put, MatwingsVenus™ (Xiaowu ™) works much like an experienced "protein architect":

Understanding the "Syntax" of Enzymes

Through its self-developed protein language model, the platform can understand from billions of natural sequences which sites are highly conserved and which regions are plastic, thus knowing "where to move and where not."

Exhaustively searching for optimal solutions in the virtual world

By combining structural information and evolutionary coupling, MatwingsVenus™ ™ can perform parallel evaluations of thousands of mutation combinations on a computer, directly targeting the most promising designs that simultaneously enhance activity, stability, and selectivity. This process compresses the "evolutionary path" that traditional methods would take years to complete, down to weeks or even days.

Experimental feedback makes AI smarter and smarter

Data from each wet experiment is fed back into the model, allowing MatwingsVenus's™ ™ design capabilities to continuously evolve. Data from multiple real industrial enzyme transformation projects show that the platform's mutant positivity rate is much higher than that of traditional random mutation libraries, allowing the acquisition of molecules that meet industrial requirements in one go.

For example, a key transaminase used in pharmaceutical intermediate synthesis has extremely low natural enzyme activity and poor thermal stability. With the help of the MatwingsVenus™ ™ platform, Matwings Technology's R&D team increased enzyme activity by over 200 times and half-life by nearly 30 times in less than two months by coordinating design at just a few key sites—a speed that was almost unimaginable in the past.

More importantly, MatwingsVenus™ is ™ not a closed black box, but a movable and scalable technological foundation. Whether in food, agriculture, biomedicine, or enzyme catalysis involved in biodegradable materials and key carbon neutrality pathways, it can quickly enter and provide customized solutions that go beyond traditional trial-and-error models.

3. The next stop in enzyme engineering: reshaping the biomanufacturing model

The Era of Intelligent Enzyme Engineering

If we compare the entire biomanufacturing industry to a car, enzymes are the 'spark plugs' of the engine. Without high-performance enzymes, no matter how ingenious the metabolic pathway design or how precise the fermentation process, it is difficult to unlock its true commercial value.

Today, enzyme engineering is standing at a singular point: with the integration of AI protein design, ultra-high-throughput screening, and automated experimental platforms, the threshold and cycle for modifying an enzyme are decreasing exponentially. This means——

More key raw materials for chiral drugs and derivatives of natural products, which were previously bottlenecked, can now be synthesized in a greener way much faster;

Artificial enzyme-catalyzed pathways using CO₂ as a raw material are no longer just theoretical concepts in papers;

Even everyday products like detergents, textiles, and food will become more sustainable thanks to more efficient and milder enzymes.

In the hands of cutting-edge explorers like Shanghai Matwings Technology, platforms such as MatwingsVenus™ (Xiaowu™) are transforming enzyme engineering from a highly experience-dependent 'craft' into a replicable and predictable engineering science.

Next time you pick up that piece of enzyme-catalyzed bread, or wear clothes dyed with biological methods, consider this: the tiny enzyme molecule behind it may not be a direct gift from nature, but rather a 'super enzyme' evolved specifically for you by a team of protein designers with the aid of AI.

Making catalysis more precise, making manufacturing greener.

The intelligent era of enzyme engineering has arrived.