How to Achieve Enzyme Enantioselective Modification?

Introduction

In the production of chiral drugs, fine chemicals, and agricultural chemicals, an enzyme’s enantioselectivity directly determines the optical purity and biological activity of the product. This is one of the main bottlenecks when using biocatalysis to replace chemical synthesis. However, the enantioselectivity of natural enzymes often doesn’t meet industrial demands. How to precisely control an enzyme’s chiral recognition through protein engineering is a key topic in enzyme engineering. This article systematically breaks down the core principles of modifying enzyme enantioselectivity, the four main methods, and the major challenges, and also introduces how Shanghai Matwings Technology’s MatwingsVenus™ (Xiaowu™) protein platform supports research and modification of chiral enzymes through directed evolution design.

1. What is Enzyme Enantioselectivity

Enantioselectivity refers to an enzyme’s ability to preferentially recognize and produce one of a pair of enantiomers during a catalytic reaction. It is an important type of stereoselectivity. The e.e. (enantiomeric excess) value measures the proportion of the dominant enantiomer in the reaction mixture and is related to the reaction conversion. The E value (enantioselectivity factor), which is the ratio of the rate constants for the two enantiomers, reflects the inherent chiral recognition ability of the enzyme itself. For industrial applications, an E value greater than 100 is usually required to obtain chiral products with more than 99% e.e. at high conversion rates.

For industrial use, an enzyme’s enantioselectivity directly affects product value and safety. In chiral drugs, one enantiomer often has drastically different pharmacological effects and side effects, making a high e.e. a core quality standard. For fine chemicals like chiral alcohols, amines, and acids, optical purity directly impacts the performance and cost of downstream products. The biological activity and sensory quality of chiral pesticides and fragrances also highly depend on enantiomeric purity.

However, natural enzymes are often “optimized” for physiological metabolism, so their enantioselectivity toward artificial substrates is generally insufficient, making them difficult to use directly for industrial production. Therefore, modifying enzyme enantioselectivity has become a key technological direction in the field of chiral biocatalysis.

2. The Four Main Methods for Modifying Enzyme Enantioselectivity

  A Comparison of Four Types of Renovation Strategies

Unlike protein stability engineering, the core challenge of modifying enantioselectivity lies in the fact that chiral recognition usually depends on fine-tuning the nanoscale stereochemistry of the active site—just changing the orientation of a single amino acid side chain can completely flip the chirality of the product. Based on this feature, four main types of modification strategies have gradually developed in the field, each with its own focus.

2.1 Structure-guided: Remodeling the stereochemistry of the active pocket

Structure-guided modification strategies rely on the enzyme’s 3D structure and catalytic mechanism. By precisely altering key residues around the substrate-binding pocket and the catalytic center, the spatial and chemical environment of the active site can be reshaped, thereby controlling enantioselectivity.

Modification strategies targeting chiral recognition are quite distinctive and mainly include: controlling steric hindrance in the active pocket—introducing bulky side-chain amino acids (like Trp or Phe) on one side of the pocket to force the substrate to enter only in a specific enantiomeric configuration; fine-tuning the chiral environment of the catalytic center—changing the hydrogen bonding network or spatial orientation near catalytic residues to influence the stability of the reaction transition state; modifying the substrate-binding channel—adjusting the stereochemistry or electrostatic distribution at the entrance of the active site so the substrate enters in a particular orientation; asymmetrically shaping hydrophobic pockets—altering the hydrophobic/polar distribution on both sides of the pocket to guide substrate orientation.

The advantage of this approach is that it requires fewer mutants and has a clear mechanism. However, it assumes you have high-quality 3D structures and a clear understanding of the catalytic mechanism, and structure-guided design often struggles to address allosteric effects distant from the active site.

2.2 Evolution-guided: Screening chiral diversity in random mutant libraries

Evolution-guided strategies generate diverse libraries through random mutations, then rely on high-throughput or highly sensitive chiral screening methods to enrich variants with improved enantioselectivity, gradually accumulating beneficial mutations over multiple rounds.

The main bottleneck here isn’t creating the library but in screening. Chiral analysis methods (like chiral HPLC, chiral GC, or chiral fluorescent probes) typically have lower throughput and higher cost than activity screening. This makes directed evolution for enantioselectivity more challenging than for stability modification—no matter how large the library, if you can’t measure it, it’s pointless. To tackle this, the field has developed several alternative strategies: colorimetric chiral indicators, host growth-based complementary screening, fluorescent probe-based chiral detection, and fluorescence-activated droplet sorting (FADS), which can enable ultra-high-throughput screening of enantioselectivity in some systems, greatly improving efficiency.

The advantage of evolution-guided approaches is that they don’t require structural information and have a large exploration space, potentially discovering unexpected modification sites. The challenges are high screening costs, long timelines, and strong dependency on the sensitivity and throughput of the screening methods.

2.3 Computational Assistance: Semi-Rational Focused Mutation Strategy

Semi-rational design with computational assistance is one of the most widely used strategies in stereoselective modification. The approach usually starts by using molecular docking, molecular dynamics simulations, quantum chemical calculations, or conservation analysis to pinpoint a set of key sites most likely to affect stereoselectivity. Then, these sites undergo saturation mutagenesis or limited combinations to create a focused library, which is subsequently screened.

Sequence conservation and correlation analysis are commonly used in semi-rational design. By aligning homologous sequences with known stereoselectivity, residues associated with chiral preference differences can be identified for targeted mutational validation.

Semi-rational design balances discovery potential with screening efficiency, making it the most commonly used paradigm in the industry for stereoselective modification. However, its effectiveness highly depends on the accuracy of computational predictions, as prediction quality directly affects the success rate of modifications.

2.4 AI Empowerment: Inferring Selective Mechanisms and Combination Optimization

In recent years, deep learning technologies, represented by protein language models (pLMs), have been expanding the boundaries of enzyme stereoselective modification. By pretraining on billions of protein sequences, pLMs can learn the deep mapping between sequence, structure, and function. Combined with structural information, they help predict which sites control substrate binding patterns and stereoselectivity, and further design multi-site combination optimization plans.

By 2026, multiple cutting-edge studies validated the effectiveness of AI methods from different perspectives: pLM-based methods enhanced or even reversed stereoselectivity in dehydrogenase families; AI-guided redesign combined with directed evolution significantly improved oxidoreductase modifications; and multi-scale machine learning/molecular mechanics (ML/MM) frameworks enabled quantitative prediction of stereoselectivity in cyclases.

AI methods offer unique value in stereoselective modification in three ways: first, remote site identification — not limited to areas around the active site, AI can find allosteric sites and distal residues that regulate stereoselectivity; second, modeling combination effects — multi-site synergy is key to improving chiral selectivity, and AI can help predict site combinations with positive synergy, overcoming the activity-selectivity trade-off; third, reducing screening dependency — computational pre-screening can narrow experimental candidates from thousands to tens, especially useful in scenarios where chiral screening throughput is limited.

💡 Quick Method Choice Guide: If you have a high-quality 3D structure and the e.e. improvement requirement is <20% → structure-guided design; if no structure/mechanism is clear but high-throughput chiral screening is available → evolution-guided screening; if screening throughput is limited and significant improvement is needed → semi-rational design or AI pre-screening, then validate using a focused library.

3. Top 5 FAQs on Enzyme Enantioselectivity Engineering

Q1: Can we do enantioselectivity engineering without crystal structures?

Yes. Traditional structure-guided approaches require 3D structures, but directed evolution and AI methods based on protein language models can start engineering using only sequences. AI can predict key sites that might affect chiral recognition from sequences, which can be used to build focused libraries, and combined with experimental screening for stepwise optimization. This is a common route for structure-free enzyme engineering.

Q2: Rational design or directed evolution—which should we choose?

Simple guideline: if you have a high-quality structure and the e.e. improvement is small (<20%), go for rational/semi-rational design; for larger improvements and if high-throughput screening is available, choose directed evolution; if screening throughput is limited, go for semi-rational design or AI pre-screen with small library validation. Currently, AI-assisted semi-rational design is one of the most efficient paradigms.

Q3: Will improving enantioselectivity always reduce activity?

Not necessarily, but the activity-selectivity trade-off is common. Tightening the active site too much may enhance chiral recognition but can also block substrate entry or product release, reducing catalytic efficiency. However, by wisely choosing mutation sites (e.g., tunnel residues away from the active site, distal allosteric sites), selectivity can often be improved without losing activity. AI methods have unique advantages here.

Q4: How many rounds does enantioselectivity engineering usually take?

It depends on the gap between the starting and target e.e. For boosting from 60% e.e. to 90% e.e., 1–2 rounds of focused library validation via semi-rational design is usually enough; to go from 60% to >99%, 3–5 rounds of iteration are often required. AI-assisted design can achieve large improvements in one step by combining mutations, reducing the number of iterations.

Q5: What if the chiral screening throughput is not enough?

This is the most common bottleneck in enantioselectivity engineering. It can be mitigated by: ① using AI pre-screening to significantly reduce library size and focus resources on the most valuable candidates; ② designing indirect screening methods based on color, fluorescence, or growth phenotype to indirectly reflect enantioselectivity; ③ using semi-rational design to build a small, high-quality focused library.

4. Four Core Challenges in Enzyme Enantioselectivity Engineering

 Schematic diagram illustrating the key challenges of the refurbishment

Despite continuous advances in methods, enzyme enantiometry selective modification still faces multiple bottlenecks in practical research.

Challenge 1: Unclear chiral recognition mechanisms and difficulty in selecting targets. The mechanisms by which enantioselectivity determines many enzymes remain unclear—is it the steric hindrance of active sites, hydrogen bond networks, or the conformational dynamics of substrate binding channels at play? Unclear mechanics lead to heavy experience in choosing modification targets, resulting in low hit rates.

Challenge 2: Chiral screening flux is low, limiting directed evolution. The core bottleneck of enantioselective modification is the screening method. Traditional chirality analysis (HPLC, GC, etc.) typically has low throughput and high cost, making it difficult to support large-scale library screening. Although new technologies such as fluorescence-activated droplet sorting (FADS) are partially alleviating this bottleneck, their applicability remains limited.

Challenge 3: Activity-Selectivity Trade-off. A common dilemma in enantioselective modification is that increasing selectivity is often accompanied by decreased catalytic activity. Although overtightening the active site pockets can improve chiral recognition, it may also hinder substrate entry or product release, leading to reduced catalytic efficiency. How to achieve the dual optimization of "high activity and high selectivity" is a core challenge in the field.

Challenge 4: Combination effects are hard to predict. Single-point mutations have limited selectivity improvements; truly significant improvements often require the synergy of multiple mutations. However, the upper-effect of multi-site combinations is extremely complex, and traditional methods struggle to predict which locus combinations will produce positive synergy, resulting in high trial-and-error costs.

5. MatwingsVenus™ (Xiaowu ™): An AI-driven directional evolution design platform

Facing the unique challenge of selective enantiometry modification, Shanghai Matwings Technology independently developed the MatwingsVenus™ (Xiaowu ™) conversational protein R&D agent, which, relying on its proprietary protein large model and directed evolution design capabilities, provides a support pathway for chiral enzyme modification from site prediction to experimental validation. Unlike general stability modification, mapping selective modification requires higher requirements for chirality recognition mechanisms, locus spatial relationships, and combinatorial synergy. The platform provides targeted support from multiple dimensions:

Wide-area scanning of distant and allele sites. A challenge in enantioselective modification is that key sites are often not limited to the active center; substrate binding channels, surface loop regions, and even distal allosteric sites can all affect chiral recognition. Based on the global sequence characterization capabilities of the Venus series protein large models, the platform can perform wide-area mutation scanning of entire sequences, not limited to residues around active sites, helping researchers discover non-classical sites that affect chiral selectivity and overcoming the blind spots of structure-oriented design.

Analysis of the synergistic effect of multiple-site combinations. Significantly improving chiral selectivity usually depends on the synergy of multiple mutations, but traditional methods struggle to predict positive combinations. The platform leverages the sequence-function mapping capability of large models to analyze candidate sites in combination, helping to identify mutation combinations with positive synergistic effects and reducing the wasted screening effort from blind library construction, which is especially suited for scenarios where chiral screening throughput is limited.

Multi-round iterative closed-loop optimization. Chiral selectivity modifications often require multiple iterations to improve from low e.e. to industrial application standards. The platform integrates computational design with automated wet experiments. High-value sites identified in one round of predictions can be converted into experimental plans with one click. After expression and functional testing in an automated shared lab, experimental results are automatically fed back to guide the next round of design, creating an iterative closed loop of "predict-experiment-feedback-reoptimize."

Conversational interaction, lowering the barrier to entry. The platform centers on an AI agent integrating over 200 protein design tools and more than 30 expert-tuned skills. Users only need to describe their chiral modification goals and constraints in natural language. The agent automatically orchestrates the relevant analysis toolchains, returning site-level candidate suggestions and functional interpretations. No command-line operations or GPU setup is needed, so even experimental researchers can quickly start modification projects.

The protein ai platform has completed multiple applied projects in innovative pharmaceuticals and synthetic biology, including PET-degrading enzymes, highly active alkaline phosphatase, and other proteins that have reached industrial stages. It covers various modification scenarios such as industrial enzymes, pharmaceutical enzymes, and functional proteins, supporting research exploration and iterative optimization for complex modifications like chiral enzymes.

Conclusion:

Upgrading from trial-and-error screening to precise design paradigms. From structure-guided "one-by-one testing" to evolution-guided "random screening," and now to AI-driven intelligent design, the evolution of methods for enzyme chiral selectivity modification has always focused on two goals: higher modification efficiency and greater selectivity improvement. Today, AI is breaking the traditional bottleneck of "long modification cycles due to insufficient screening throughput" by using computational pre-screening to narrow the search space and focusing limited screening resources on the most valuable candidates.

Shanghai Matwings Technology's Matwings Venus™ (Xiaowu™) platform is driving this paradigm shift: it centers on AI-driven directed evolution and protein design, connects upward to billion-scale sequence databases and protein large models, and integrates downward with functional design, automated experimental validation, and wet-dry closed-loop iteration. This compresses the long traditional "library-building, screening, iteration" R&D cycle into an efficient "predict-validate-optimize" closed loop.

The next breakthrough in chiral biocatalysis belongs to researchers who can deeply integrate AI intelligent design with experimental validation. Matwings Venus™ (Xiaowu™) is providing the unified capability foundation for such research.