發布日期：2025-09-23

• 酵素進化工程的核心概念與技術前沿

酵素，又稱為酶，是生物體內最精密的分子，透過高度專一催化效率推動化學反應。在過去數十年間，酵素工程的進展依賴「定向演化」（directed evolution）或「非定向演化」(non-directed evolution) ── 透過定向設計半隨機或是隨機自然的突變方式篩選，逐步累積有利突變，最終獲得性能優化的酵素。這一方法奠定了現代酵素工程的基石，並已在藥物合成、工業生物技術與綠色能源等領域展現巨大價值。然而定向與非定向演化方式受限於實驗篩選規模與數據積累效率，通常需要漫長的週期和高昂的成本。

隨著結構生物學、計算模擬、合成生物學與人工智慧的融合，研究者不再僅依賴半隨機或自然篩選，而是逐步轉向「設計導向」的策略。計算機輔助蛋白質設計、分子動力學模擬以及機器學習模型等平台（如 AlphaFold、Rosetta、AutoDock Vina），使科學家能夠在虛擬空間中探索龐大的序列與結構空間，預測穩定性、溶解性與基質專一性，並針對性地提出候選突變方案。進一步，自動化實驗工作站與微流控系統的工藝進步，促使連續進化系統與高通量篩選更加普及，則讓「設計—製造—測試—學習」的經典合成生物流程循環成為全自動化的運行，大幅縮短了酵素優化的週期以及降低人力資源的浪費。今日的酵素進化工程，正處於從自然演化啟發的隨機探索，走向人工智慧輔助的精準設計的關鍵階段。

• 高通量篩選（High-throughput screening）

在酵素進化工程中，篩選（selection/screening）往往是最大的瓶頸。自動化實驗工作站依靠 96 孔、 384 或1536 孔板，處理數千變體；而微流控系統將反應縮小至微米級液滴，單一液滴相當於一個獨立的反應器，能在短時間內處理數百萬甚至上億的酶變體。這兩種系統各有其適合的應用場景(如 Table 1)，讓高通量篩選成為可能，為定向進化和連續進化提供了強大的篩選引擎。在蛋白質進化工程中有許多可以使用的工具，我們將其整理在圖二。

Core Concepts and Technological Frontiers of Enzyme Evolution Engineering

Enzymes, also known as catalysts, are among the most precise molecules in living organisms, driving chemical reactions with high specificity and efficiency. Over the past several decades, advances in enzyme engineering have relied on directed evolution or non-directed evolution — strategies that employ semi-random, targeted, or naturally random mutations combined with screening to gradually accumulate beneficial changes, ultimately yielding optimized enzymes. These approaches have laid the foundation for modern enzyme engineering and have demonstrated significant value in drug synthesis, industrial biotechnology, and green energy. However, both directed and non-directed evolution remain constrained by the scale of experimental screening and the efficiency of data accumulation, often requiring often requiring lengthy cycles and high costs.

With the integration of structural biology, computational modeling, synthetic biology, and artificial intelligence, researchers are no longer dependent solely on random screening and are increasingly turning to design-driven strategies. Computational protein design platforms , molecular dynamics simulations, and machine learning models (such as AlphaFold, Rosetta, and AutoDock Vina) enable scientists to explore vast sequence and structural spaces in silico, predict stability, solubility, and substrate specificity, and propose rational mutation candidates. Furthermore, advances in automated experimental workstations and microfluidics have made continuous evolution platforms and high-throughput screening more accessible, driving the automation of the classical Design–Build–Test–Learn (DBTL) cycle in synthetic biology, thereby significantly shortening the timescale for enzyme optimization. Today, enzyme evolution engineering is at a pivotal stage, transitioning from random exploration inspired by natural evolution toward precise, AI-assisted design.

Enzyme Evolution Engineering Methods

The core challenge of enzyme engineering lies in converting protein sequences into desired protein functions. Natural enzymes often lack sufficient activity, specificity, or stability to meet industrial and pharmaceutical demands, necessitating the application of evolutionary engineering at multiple levels to enhance performance. These methods (Fig. 1) range from random mutagenesis and screening that mimic natural evolution, to structure-informed rational design, and further to AI-driven prediction and design, forming a comprehensive technological spectrum that spans natural and artificial processes, as well as empirical and algorithmic approaches.

High-Throughput Screening 

In enzyme evolution engineering, selection and screening often constitute the major bottlenecks. Automated experimental workstations, operating with 96-, 384-, or 1536-well plates, can process thousands of variants, whereas microfluidics miniaturizes reactions into micrometer-sized droplets, with each droplet acting as an independent reactor capable of screening millions to hundreds of millions of enzyme variants in a short timeframe. Each system has distinct application scenarios (Table. 1), together enabling high-throughput screening and providing a powerful selection engine for directed enzyme evolution. Numerous tools are available for protein evolution engineering, which we summarize in Fig. 2.

Enzyme engineering represents the convergence of molecular biology, computational modeling, and high-throughput experimentation, demonstrating transformative potential in biotechnology, medicine, and sustainable chemistry. By integrating de novo design, AI-assisted prediction, and directed evolution, researchers can not only accelerate the discovery of highly active and specific catalysts but also expand the functional boundaries of enzymes. The cross-disciplinary nature of this field underscores its significance: advances in enzyme engineering drive innovations in drug development, industrial biocatalysis, and environmental solutions, highlighting its role as a cornerstone of modern synthetic biology.