AI+定向进化赋能蛋白改造及优化

doi:10.12211/2096-8280.2025-044

合成生物学 ›› 2025, Vol. 6 ›› Issue (3): 617-635.DOI: 10.12211/2096-8280.2025-044

AI+定向进化赋能蛋白改造及优化

宋成治¹(), 林一瀚¹^,²^,³()

^1.北京大学定量生物学中心，北京大学-清华大学生命科学联合中心，北京大学前沿学科交叉研究院，北京 100871
^2.北京大学教育部细胞增殖与分化重点实验室，北京大学生命科学学院，北京 100871
^3.北京大学成都前沿交叉生物技术研究院，四川成都 610213

收稿日期:2025-05-12 修回日期:2025-06-06 出版日期:2025-06-30 发布日期:2025-06-27
通讯作者: 林一瀚
作者简介:宋成治（1998—），男，博士研究生。研究方向包括系统生物学、合成生物学、生物物理。 E-mail：czsong@stu.pku.edu.cn
基金资助:
国家重点研发计划(2020YFA0906900)

AI-enabled directed evolution for protein engineering and optimization

SONG Chengzhi¹(), LIN Yihan¹^,²^,³()

^1.Center for Quantitative Biology，Peking-Tsinghua Center for Life Sciences，Academy for Advanced Interdisciplinary Studies，Peking University，Beijing 100871，China
^2.The MOE Key Laboratory of Cell Proliferation and Differentiation，School of Life Sciences，Peking University，Beijing 100871，China
^3.Chengdu Academy for Advanced Interdisciplinary Biotechnologies，Peking University，Chengdu 610213，Sichuan，China

Received:2025-05-12 Revised:2025-06-06 Online:2025-06-30 Published:2025-06-27
Contact: LIN Yihan

摘要/Abstract

摘要：

定向进化是合成生物学领域的核心底层技术之一。通过在实验室中模拟自然界发生的进化过程，定向进化利用功能筛选从大量的突变序列文库中不断获得性能提升的蛋白序列，帮助实现野生型蛋白难以实现的功能。近年来不断发展的机器学习、蛋白语言模型等人工智能（artificial intelligence， AI）方法进一步拓展了该技术的使用场景和工作效率，帮助其在酶、抗体、生物传感器等的改造中取得优异表现。本文总结了传统定向进化在突变文库构建和功能筛选过程中使用的典型策略，并对近年来开发的高效连续定向进化平台进行介绍，进一步对定向进化技术存在的序列空间有限、容易陷入局部最优等一系列问题进行探讨。快速迭代的机器学习模型与定向进化相结合，一方面能够缓解序列空间的探索局限性，另一方面能够从起始序列设计、中间文库优化、功能信息提取等多个维度对定向进化的实验流程进行完善，帮助实现更加高效的蛋白改造尝试。为明确定向进化结合机器学习的应用潜力，本文重点展示了机器学习辅助定向进化的代表案例。最后，简要探讨了该领域的潜在挑战和未来发展方向。

关键词: 定向进化, 机器学习, 蛋白改造, 蛋白语言模型, 合成生物学

Abstract:

Directed evolution is one of the core enabling technologies in synthetic biology. By recapitulating evolutionary processes that occur in nature within the laboratories, directed evolution employs functional screening to continually isolate variants with improved performance from large mutant libraries for functions that are difficult to achieve with wild-type proteins. In recent years, rapidly advancing artificial intelligence (AI) approaches—such as machine learning and protein language models—have further expanded both the range of applications and the operational efficiency of directed evolution, yielding unprecedented achievements in the engineering of enzymes, antibodies, biosensors, and more. In this review, we first outline classic strategies and emerging techniques for mutagenesis and functional selection in traditional directed evolution, followed by an in-depth examination of various continuous directed evolution systems. We highlight common limitations of directed evolution, emphasizing issues such as constrained search space and susceptibility to local optima. Combining rapidly iterated AI methods with directed evolution offers promising solutions to these challenges. Protein language models, in particular, leverage learned patterns from experimental variants alongside fundamental protein properties, providing superior predictive accuracy for unexplored mutants and facilitating the extrapolation of sequence-function relationships to broader sequence space. AI-based methods enhance directed evolution workflows from multiple perspectives. De novo protein design and unsupervised protein language models aid in generating functional starting sequences with targeted sequence diversity. Machine learning models trained on experimental data enable the construction of optimized mutant libraries tailored for subsequent selection rounds. Additionally, models derived from statistical physics and dynamical systems help extract detailed functional information from data acquired across multiple selection rounds. Collectively, these machine learning approaches significantly enhance the overall efficiency of directed evolution. To illustrate the transformative potential of machine learning-assisted directed evolution, we discuss exemplary cases of protein function improvement and modification. Lastly, we briefly address ongoing challenges and future directions in this rapidly evolving and promising research area.

Key words: directed evolution, machine learning, protein engineering, protein language model, synthetic biology

中图分类号:

Q816

宋成治, 林一瀚. AI+定向进化赋能蛋白改造及优化[J]. 合成生物学, 2025, 6(3): 617-635.

SONG Chengzhi, LIN Yihan. AI-enabled directed evolution for protein engineering and optimization[J]. Synthetic Biology Journal, 2025, 6(3): 617-635.

图/表 4

参考文献 60

61	WEINREICH D M, DELANEY N F, DEPRISTO M A, et al. Darwinian evolution can follow only very few mutational paths to fitter proteins[J]. Science, 2006, 312(5770): 111-114.
62	PODGORNAIA A I, LAUB M T. Pervasive degeneracy and epistasis in a protein-protein interface[J]. Science, 2015, 347(6222): 673-677.
63	FOX R, ROY A, GOVINDARAJAN S, et al. Optimizing the search algorithm for protein engineering by directed evolution[J]. Protein Engineering Design and Selection, 2003, 16(8): 589-597.
64	FOX R J, DAVIS S C, MUNDORFF E C, et al. Improving catalytic function by ProSAR-driven enzyme evolution[J]. Nature Biotechnology, 2007, 25(3): 338-344.
65	ROMERO P A, KRAUSE A, ARNOLD F H. Navigating the protein fitness landscape with Gaussian processes[J]. Proceedings of the National Academy of Sciences of the United States of America, 2013, 110(3): E193-E201.
66	OTWINOWSKI J, PLOTKIN J B. Inferring fitness landscapes by regression produces biased estimates of epistasis[J]. Proceedings of the National Academy of Sciences of the United States of America, 2014, 111(22): E2301-E2309.
67	OFER D, BRANDES N, LINIAL M. The language of proteins: NLP, machine learning & protein sequences[J]. Computational and Structural Biotechnology Journal, 2021, 19: 1750-1758.
68	FERRUZ N, HÖCKER B. Controllable protein design with language models[J]. Nature Machine Intelligence, 2022, 4(6): 521-532.
69	ASGARI E, MOFRAD M R K. Continuous distributed representation of biological sequences for deep proteomics and genomics[J]. PLoS One, 2015, 10(11): e0141287.
70	HEINZINGER M, ELNAGGAR A, WANG Y, et al. Modeling aspects of the language of life through transfer-learning protein sequences[J]. BMC Bioinformatics, 2019, 20(1): 723.
71	ALLEY E C, KHIMULYA G, BISWAS S, et al. Unified rational protein engineering with sequence-based deep representation learning[J]. Nature Methods, 2019, 16(12): 1315-1322.
72	RAO R, BHATTACHARYA N, THOMAS N, et al. Evaluating protein transfer learning with TAPE[C]//Advances in Neural Information Processing Systems 32 (NeurIPS 2019), 2019, 32: 9689-9701 [2025-06-03]. .
73	MADANI A, KRAUSE B, GREENE E R, et al. Large language models generate functional protein sequences across diverse families[J]. Nature Biotechnology, 2023, 41(8): 1099-1106.
74	MADANI A, MCCANN B, NAIK N, et al. ProGen: language modeling for protein generation[EB/OL]. arXiv, 2020: 2004.03497. (2020-03-08)[2025-06-03]. .
75	RIVES A, MEIER J, SERCU T, et al. Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences[J]. Proceedings of the National Academy of Sciences of the United States of America, 2021, 118(15): e2016239118.
76	MEIER J, RAO R, VERKUIL R, et al. Language models enable zero-shot prediction of the effects of mutations on protein function[C/OL]//Advances in Neural Information Processing Systems 34 (NeurIPS 2021), 2021, 34: 29287-29303 [2025-06-03]. .
77	RAO R M, LIU J, VERKUIL R, et al. MSA transformer[C]//Proceedings of the 38th International Conference on Machine Learning, PMLR, 2021, 139: 8844-8856 [2025-06-03]. .
78	BRANDES N, OFER D, PELEG Y, et al. ProteinBERT: a universal deep-learning model of protein sequence and function[J]. Bioinformatics, 2022, 38(8): 2102-2110.
79	FERRUZ N, SCHMIDT S, HÖCKER B. ProtGPT2 is a deep unsupervised language model for protein design[J]. Nature Communications, 2022, 13: 4348.
80	ELNAGGAR A, HEINZINGER M, DALLAGO C, et al. ProtTrans: toward understanding the language of life through self-supervised learning[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(10): 7112-7127.
81	LIN Z M, AKIN H, RAO R, et al. Evolutionary-scale prediction of atomic-level protein structure with a language model[J]. Science, 2023, 379(6637): 1123-1130.
82	NIJKAMP E, RUFFOLO J A, WEINSTEIN E N, et al. ProGen2: exploring the boundaries of protein language models[J]. Cell Systems, 2023, 14(11): 968-978.e3.
83	ELNAGGAR A, ESSAM H, SALAH-ELDIN W, et al. Ankh: optimized protein language model unlocks general-purpose modelling[EB/OL]. arXiv, 2023: 2301.06568.(2023-01-16)[2025-06-03]. .
84	TRUONG T F JR, BEPLER T. PoET: a generative model of protein families as sequences-of-sequences[C/OL]//Advances in Neural Information Processing Systems 36 (NeurIPS 2023), 2023: 2306.06156[2025-06-03]. .
1	LERNER S A, WU T T, LIN E C. Evolution of a catabolic pathway in bacteria[J]. Science, 1964, 146(3649): 1313-1315.
2	HALL B G. Experimental evolution of a new enzymatic function. Kinetic analysis of the ancestral (ebg ^o) and evolved (ebg ⁺) enzymes[J]. Journal of Molecular Biology, 1976, 107(1): 71-84.
3	LENUG D W, CHEN E, GOEDDEL D V. A method for random mutagenesis of a defined DNA segment using a modified polymerase chain reaction[J]. Technique JMCMB, 1989, 1: 11-15.
4	CHEN K Q, ARNOLD F H. Enzyme engineering for nonaqueous solvents: random mutagenesis to enhance activity of subtilisin E in polar organic media[J]. Nature Biotechnology, 1991, 9(11): 1073-1077.
5	KLENK C, SCRIVENS M, NIEDERER A, et al. A Vaccinia-based system for directed evolution of GPCRs in mammalian cells[J]. Nature Communications, 2023, 14: 1770.
6	HUFFMAN M A, FRYSZKOWSKA A, ALVIZO O, et al. Design of an in vitro biocatalytic cascade for the manufacture of islatravir[J]. Science, 2019, 366(6470): 1255-1259.
7	TOURNIER V, TOPHAM C M, GILLES A, et al. An engineered PET depolymerase to break down and recycle plastic bottles[J]. Nature, 2020, 580(7802): 216-219.
8	SARAI N S, FULTON T J, O’MEARA R L, et al. Directed evolution of enzymatic silicon-carbon bond cleavage in siloxanes[J]. Science, 2024, 383(6681): 438-443.
9	RAPPAZZO C G, TSE L V, KAKU C I, et al. Broad and potent activity against SARS-like viruses by an engineered human monoclonal antibody[J]. Science, 2021, 371(6531): 823-829.
10	BANACH B B, PLETNEV S, OLIA A S, et al. Antibody-directed evolution reveals a mechanism for enhanced neutralization at the HIV-1 fusion peptide site[J]. Nature Communications, 2023, 14: 7593.
11	TABEBORDBAR M, LAGERBORG K A, STANTON A, et al. Directed evolution of a family of AAV capsid variants enabling potent muscle-directed gene delivery across species[J]. Cell, 2021, 184(19): 4919-4938.e22.
12	LIN R, ZHOU Y T, YAN T, et al. Directed evolution of adeno-associated virus for efficient gene delivery to microglia[J]. Nature Methods, 2022, 19(8): 976-985.
13	CLARKE J, FERSHT A R. Engineered disulfide bonds as probes of the folding pathway of barnase: increasing the stability of proteins against the rate of denaturation[J]. Biochemistry, 1993, 32(16): 4322-4329.
14	REA V, KOLKMAN A J, VOTTERO E, et al. Active site substitution A82W improves the regioselectivity of steroid hydroxylation by cytochrome P450 BM3 mutants as rationalized by spin relaxation nuclear magnetic resonance studies[J]. Biochemistry, 2012, 51(3): 750-760.
85	HAYES T, RAO R, AKIN H, et al. Simulating 500 million years of evolution with a language model[J]. Science, 2025, 387(6736): 850-858.
86	YANG K K, FUSI N, LU A X. Convolutions are competitive with transformers for protein sequence pretraining[J]. Cell Systems, 2024, 15(3): 286-294.e2.
87	LV L, LIN Z Y, LI H, et al. ProLLaMA: a protein large language model for multi-task protein language processing[J]. IEEE Transactions on Artificial Intelligence, 2025, PP(99): 1-12.
88	CHEN B, CHENG X Y, LI P, et al. xTrimoPGLM: unified 100B-scale pre-trained transformer for deciphering the language of protein[EB/OL]. arXiv, 2024: 2401.06199. (2024-01-11)[2025-06-03]. .
89	SAYEED M A, TEKIN E, NADEEM M, et al. Prot42: a novel family of protein language models for target-aware protein binder generation[EB/OL]. arXiv, 2025: 2504.04453. (2025-04-06)[2025-06-03]. .
90	KELLY T, XIA S, LU J Y, et al. Unified deep learning of molecular and protein language representations with T5ProtChem[J]. Journal of Chemical Information and Modeling, 2025, 65(8): 3990-3998.
91	WANG Y H, WANG Z C, SADEH G, et al. LC-PLM: long-context protein language modeling using bidirectional mamba with shared projection layers[EB/OL]. arXiv, 2025::2411.08909. (2024-10-29)[2025-06-03]. .
92	BISWAS S, KHIMULYA G, ALLEY E C, et al. Low-N protein engineering with data-efficient deep learning[J]. Nature Methods, 2021, 18(4): 389-396.
93	SUZEK B E, WANG Y Q, HUANG H Z, et al. UniRef clusters: a comprehensive and scalable alternative for improving sequence similarity searches[J]. Bioinformatics, 2015, 31(6): 926-932.
94	RADFORD A, JOZEFOWICZ R, SUTSKEVER I. Learning to generate reviews and discovering sentiment[EB/OL]. arXiv, 2017: 1704.01444. (2017-04-05)[2025-06-03]. .
95	HSU C, VERKUIL R, LIU J, et al. Learning inverse folding from millions of predicted structures[C/OL]//Proceedings of the 39th International Conference on Machine Learning, PMLR, 2022, 162: 8946-8970 [2025-06-03]. .
96	JUMPER J, EVANS R, PRITZEL A, et al. Highly accurate protein structure prediction with AlphaFold[J]. Nature, 2021, 596(7873): 583-589.
15	MAITI A, BUFFALO C Z, SAURABH S, et al. Structural and photophysical characterization of the small ultra-red fluorescent protein[J]. Nature Communications, 2023, 14: 4155.
16	PACKER M S, LIU D R. Methods for the directed evolution of proteins[J]. Nature Reviews Genetics, 2015, 16(7): 379-394.
17	WANG Y J, XUE P, CAO M F, et al. Directed evolution: methodologies and applications[J]. Chemical Reviews, 2021, 121(20): 12384-12444.
18	ZACCOLO M, WILLIAMS D M, BROWN D M, et al. An approach to random mutagenesis of DNA using mixtures of triphosphate derivatives of nucleoside analogues[J]. Journal of Molecular Biology, 1996, 255(4): 589-603.
19	VANHERCKE T, AMPE C, TIRRY L, et al. Reducing mutational bias in random protein libraries[J]. Analytical Biochemistry, 2005, 339(1): 9-14.
20	DENNIG A, SHIVANGE A V, MARIENHAGEN J, et al. OmniChange: the sequence independent method for simultaneous site-saturation of five codons[J]. PLoS One, 2011, 6(10): e26222.
21	PÜLLMANN P, ULPINNIS C, MARILLONNET S, et al. Golden Mutagenesis: an efficient multi-site-saturation mutagenesis approach by Golden Gate cloning with automated primer design[J]. Scientific Reports, 2019, 9: 10932.
22	ZHAO H M, GIVER L, SHAO Z X, et al. Molecular evolution by staggered extension process (StEP) in vitro recombination[J]. Nature Biotechnology, 1998, 16(3): 258-261.
23	COCO W M, LEVINSON W E, CRIST M J, et al. DNA shuffling method for generating highly recombined genes and evolved enzymes[J]. Nature Biotechnology, 2001, 19(4): 354-359.
24	SIEBER V, MARTINEZ C A, ARNOLD F H. Libraries of hybrid proteins from distantly related sequences[J]. Nature Biotechnology, 2001, 19(5): 456-460.
25	BITTKER J A, LE B V, LIU J M, et al. Directed evolution of protein enzymes using nonhomologous random recombination[J]. Proceedings of the National Academy of Sciences of the United States of America, 2004, 101(18): 7011-7016.
26	GREENER A, CALLAHAN M, JERPSETH B. An efficient random mutagenesis technique using an E.coli mutator strain[J]. Molecular Biotechnology, 1997, 7(2): 189-195.
97	NOTIN P, KOLLASCH A, RITTER D, et al. ProteinGym: large-scale benchmarks for protein fitness prediction and design[C/OL]//Advances in Neural Information Processing Systems 36 (NeurIPS 2023), 2023: 64331-64379 [2025-06-03]. .
98	BADRAN A H, GUZOV V M, HUAI Q, et al. Continuous evolution of Bacillus thuringiensis toxins overcomes insect resistance[J]. Nature, 2016, 533(7601): 58-63.
99	GLÖGL M, KRISHNAKUMAR A, RAGOTTE R J, et al. Target-conditioned diffusion generates potent TNFR superfamily antagonists and agonists[J]. Science, 2024, 386(6726): 1154-1161.
100	VÁZQUEZ TORRES S, BENARD VALLE M, MACKESSY S P, et al. De novo designed proteins neutralize lethal snake venom toxins[J]. Nature, 2025, 639(8053): 225-231.
101	YEH A H W, NORN C, KIPNIS Y, et al. De novo design of luciferases using deep learning[J]. Nature, 2023, 614(7949): 774-780.
102	KIPNIS Y, CHAIB A O, VOROBIEVA A A, et al. Design and optimization of enzymatic activity in a de novo β-barrel scaffold[J]. Protein Science, 2022, 31(11): e4405.
103	DING K, CHIN M, ZHAO Y L, et al. Machine learning-guided co-optimization of fitness and diversity facilitates combinatorial library design in enzyme engineering[J]. Nature Communications, 2024, 15: 6392.
104	FRAM B, SU Y, TRUEBRIDGE I, et al. Simultaneous enhancement of multiple functional properties using evolution-informed protein design[J]. Nature Communications, 2024, 15: 5141.
105	WU Z, JENNIFER KAN S B, LEWIS R D, et al. Machine learning-assisted directed protein evolution with combinatorial libraries[J]. Proceedings of the National Academy of Sciences of the United States of America, 2019, 116(18): 8852-8858.
106	WU N C, DAI L, OLSON C A, et al. Adaptation in protein fitness landscapes is facilitated by indirect paths[J]. eLife, 2016, 5: e16965.
107	CHU H Y, FONG J H C, THEAN D G L, et al. Accurate top protein variant discovery via low-N pick-and-validate machine learning[J]. Cell Systems, 2024, 15(2): 193-203.e6.
108	YANG J, LAL R G, BOWDEN J C, et al. Active learning-assisted directed evolution[J]. Nature Communications, 2025, 16: 714.
27	BADRAN A H, LIU D R. Development of potent in vivo mutagenesis plasmids with broad mutational spectra[J]. Nature Communications, 2015, 6: 8425.
28	MOORE C L, PAPA L J 3 RD, SHOULDERS M D. A processive protein Chimera introduces mutations across defined DNA regions in vivo [J]. Journal of the American Chemical Society, 2018, 140(37): 11560-11564.
29	RAVIKUMAR A, ARZUMANYAN G A, OBADI M K A, et al. Scalable, continuous evolution of genes at mutation rates above genomic error thresholds[J]. Cell, 2018, 175(7): 1946-1957.e13.
30	RAVIKUMAR A, ARRIETA A, LIU C C. An orthogonal DNA replication system in yeast[J]. Nature Chemical Biology, 2014, 10(3): 175-177.
31	HALPERIN S O, TOU C J, WONG E B, et al. CRISPR-guided DNA polymerases enable diversification of all nucleotides in a tunable window[J]. Nature, 2018, 560(7717): 248-252.
32	TOU C J, SCHAFFER D V, DUEBER J E. Targeted diversification in the S. cerevisiae genome with CRISPR-guided DNA polymerase I[J]. ACS Synthetic Biology, 2020, 9(7): 1911-1916.
33	ÁLVAREZ B, MENCÍA M, DE LORENZO V, et al. In vivo diversification of target genomic sites using processive base deaminase fusions blocked by dCas9[J]. Nature Communications, 2020, 11: 6436.
34	YI X, KHEY J, KAZLAUSKAS R J, et al. Plasmid hypermutation using a targeted artificial DNA replisome[J]. Science Advances, 2021, 7(29): eabg8712.
35	CROOK N, ABATEMARCO J, SUN J, et al. In vivo continuous evolution of genes and pathways in yeast[J]. Nature Communications, 2016, 7: 13051.
36	CARR P A, WANG H H, STERLING B, et al. Enhanced multiplex genome engineering through co-operative oligonucleotide co-selection[J]. Nucleic Acids Research, 2012, 40(17): e132.
37	LEWIS J C, ARNOLD F H. Catalysts on demand: selective oxidations by laboratory-evolved cytochrome P450 BM3[J]. Chimia, 2009, 63(6): 309.
38	COELHO P S, BRUSTAD E M, KANNAN A, et al. Olefin cyclopropanation via carbene transfer catalyzed by engineered cytochrome P450 enzymes[J]. Science, 2013, 339(6117): 307-310.
109	ZHOU Z Y, ZHANG L, YU Y X, et al. Enhancing efficiency of protein language models with minimal wet-lab data through few-shot learning[J]. Nature Communications, 2024, 15: 5566.
110	WITTMANN B J, YUE Y S, ARNOLD F H. Informed training set design enables efficient machine learning-assisted directed protein evolution[J]. Cell Systems, 2021, 12(11): 1026-1045.e7.
111	THURONYI B W, KOBLAN L W, LEVY J M, et al. Continuous evolution of base editors with expanded target compatibility and improved activity[J]. Nature Biotechnology, 2019, 37(9): 1070-1079.
112	HU J H, MILLER S M, GEURTS M H, et al. Evolved Cas9 variants with broad PAM compatibility and high DNA specificity[J]. Nature, 2018, 556(7699): 57-63.
113	JUDGE A, SANKARAN B, HU L Y, et al. Network of epistatic interactions in an enzyme active site revealed by large-scale deep mutational scanning[J]. Proceedings of the National Academy of Sciences of the United States of America, 2024, 121(12): e2313513121.
114	OLSON C A, WU N C, SUN R. A comprehensive biophysical description of pairwise epistasis throughout an entire protein domain[J]. Current Biology, 2014, 24(22): 2643-2651.
115	LIU G, ZENG H Y, MUELLER J, et al. Antibody complementarity determining region design using high-capacity machine learning[J]. Bioinformatics, 2020, 36(7): 2126-2133.
116	FERNANDEZ-DE-COSSIO-DIAZ J, UGUZZONI G, PAGNANI A. Unsupervised inference of protein fitness landscape from deep mutational scan[J]. Molecular Biology and Evolution, 2021, 38(1): 318-328.
117	SESTA L, UGUZZONI G, FERNANDEZ-DE-COSSIO-DIAZ J, et al. AMaLa: analysis of directed evolution experiments via annealed mutational approximated landscape[J]. International Journal of Molecular Sciences, 2021, 22(20): 10908.
118	SHEN M W, ZHAO K T, LIU D R. Reconstruction of evolving gene variants and fitness from short sequencing reads[J]. Nature Chemical Biology, 2021, 17(11): 1188-1198.
119	ALVAREZ S, NARTEY C M, MERCADO N, et al. In vivo functional phenotypes from a computational epistatic model of evolution[J]. Proceedings of the National Academy of Sciences of the United States of America, 2024, 121(6): e2308895121.
120	DI BARI L, BISARDI M, COTOGNO S, et al. Emergent time scales of epistasis in protein evolution[J]. Proceedings of the National Academy of Sciences of the United States of America, 2024, 121(40): e2406807121.
121	JIANG K Y, YAN Z Q, DI BERNARDO M, et al. Rapid in silico directed evolution by a protein language model with EVOLVEpro[J]. Science, 2025, 387(6732): eadr6006.
122	LANDWEHR G M, BOGART J W, MAGALHAES C, et al. Accelerated enzyme engineering by machine-learning guided cell-free expression[J]. Nature Communications, 2025, 16: 865.
123	JIANG F, LI M C, DONG J J, et al. A general temperature-guided language model to design proteins of enhanced stability and activity[J]. Science Advances, 2024, 10(48): eadr2641.
124	HIE B L, SHANKER V R, XU D, et al. Efficient evolution of human antibodies from general protein language models[J]. Nature Biotechnology, 2024, 42(2): 275-283.
125	SHANKER V R, BRUUN T U J, HIE B L, et al. Unsupervised evolution of protein and antibody complexes with a structure-informed language model[J]. Science, 2024, 385(6704): 46-53.
126	BEDBROOK C N, YANG K K, ROBINSON J E, et al. Machine learning-guided channelrhodopsin engineering enables minimally invasive optogenetics[J]. Nature Methods, 2019, 16(11): 1176-1184.
127	UNGER E K, KELLER J P, ALTERMATT M, et al. Directed evolution of a selective and sensitive serotonin sensor via machine learning[J]. Cell, 2020, 183(7): 1986-2002.e26.
128	SAITO Y, OIKAWA M, NAKAZAWA H, et al. Machine-learning-guided mutagenesis for directed evolution of fluorescent proteins[J]. ACS Synthetic Biology, 2018, 7(9): 2014-2022.
129	CHENG X Y, CHEN B, LI P, et al. Training compute-optimal protein language models[C/OL]//Advances in Neural Information Processing Systems 37 (NeurIPS 2024), 2024, 37: 69386-69418 [2025-06-03]. .
130	LUO Y N, JIANG G D, YU T H, et al. ECNet is an evolutionary context-integrated deep learning framework for protein engineering[J]. Nature Communications, 2021, 12: 5743.
131	LI M C, KANG L Q, XIONG Y, et al. SESNet: sequence-structure feature-integrated deep learning method for data-efficient protein engineering[J]. Journal of Cheminformatics, 2023, 15(1): 12.
132	DIECKHAUS H, KUHLMAN B. Protein stability models fail to capture epistatic interactions of double point mutations[EB/OL]. bioRxiv, 2024: 2024.08.20.608844. (2024-08-20)[2025-06-03]. .
39	CHEN H Q, LIU S, PADULA S, et al. Efficient, continuous mutagenesis in human cells using a pseudo-random DNA editor[J]. Nature Biotechnology, 2020, 38(2): 165-168.
40	STIFFLER M A, HEKSTRA D R, RANGANATHAN R. Evolvability as a function of purifying selection in TEM-1 β-lactamase[J]. Cell, 2015, 160(5): 882-892.
41	STARR T N, GREANEY A J, HILTON S K, et al. Deep mutational scanning of SARS-CoV-2 receptor binding domain reveals constraints on folding and ACE2 binding[J]. Cell, 2020, 182(5): 1295-1310.e20.
42	MA L, LIN Y H. Orthogonal RNA replication enables directed evolution and Darwinian adaptation in mammalian cells[J]. Nature Chemical Biology, 2025, 21(3): 451-463.
43	COLLINS C H, LEADBETTER J R, ARNOLD F H. Dual selection enhances the signaling specificity of a variant of the quorum-sensing transcriptional activator LuxR[J]. Nature Biotechnology, 2006, 24(6): 708-712.
44	MORRISON M S, PODRACKY C J, LIU D R. The developing toolkit of continuous directed evolution[J]. Nature Chemical Biology, 2020, 16(6): 610-619.
45	MOLINA R S, RIX G, MENGISTE A A, et al. In vivo hypermutation and continuous evolution[J]. Nature Reviews Methods Primers, 2022, 2: 36.
46	ESVELT K M, CARLSON J C, LIU D R. A system for the continuous directed evolution of biomolecules[J]. Nature, 2011, 472(7344): 499-503.
47	PACKER M S, REES H A, LIU D R. Phage-assisted continuous evolution of proteases with altered substrate specificity[J]. Nature Communications, 2017, 8: 956.
48	BLUM T R, LIU H, PACKER M S, et al. Phage-assisted evolution of botulinum neurotoxin proteases with reprogrammed specificity[J]. Science, 2021, 371(6531): 803-810.
49	MILLER S M, WANG T N, RANDOLPH P B, et al. Continuous evolution of SpCas9 variants compatible with non-G PAMs[J]. Nature Biotechnology, 2020, 38(4): 471-481.
50	RICHTER M F, ZHAO K T, ETON E, et al. Phage-assisted evolution of an adenine base editor with improved Cas domain compatibility and activity[J]. Nature Biotechnology, 2020, 38(7): 883-891.
133	YU T H, BOOB A G, SINGH N, et al. In vitro continuous protein evolution empowered by machine learning and automation[J]. Cell Systems, 2023, 14(8): 633-644.
134	GELMAN S, JOHNSON B, FRESCHLIN C, et al. Biophysics-based protein language models for protein engineering[EB/OL]. bioRxiv, 2024: 2024.03.15.585128. (2024-03-15)[2025-06-03]. .
135	OLIVARES-GIL A, BARBERO-APARICIO J A, RODRÍGUEZ J J, et al. Semi-supervised prediction of protein fitness for data-driven protein engineering[J]. Journal of Cheminformatics, 2025, 17(1): 88.
136	VIG J, MADANI A, VARSHNEY L R, et al. BERTology meets biology: interpreting attention in protein language models[EB/OL]. arXiv, 2020: 2006.15222. (2020-06-26)[2025-06-03]. .
137	CHEN L, ZHANG Z H, LI Z H, et al. Learning protein fitness landscapes with deep mutational scanning data from multiple sources[J]. Cell Systems, 2023, 14(8): 706-721.e5.
51	MERCER J A M, DECARLO S J, ROY BURMAN S S, et al. Continuous evolution of compact protein degradation tags regulated by selective molecular glues[J]. Science, 2024, 383(6688): eadk4422.
52	ENGLISH J G, OLSEN R H J, LANSU K, et al. VEGAS as a platform for facile directed evolution in mammalian cells[J]. Cell, 2019, 178(3): 748-761.e17.
53	DENES C E, COLE A J, TRAN M T N, et al. The VEGAS platform is unsuitable for mammalian directed evolution[J]. ACS Synthetic Biology, 2022, 11(10): 3544-3549.
54	KIMMAN T G, SMIT E, KLEIN M R. Evidence-based biosafety: a review of the principles and effectiveness of microbiological containment measures[J]. Clinical Microbiology Reviews, 2008, 21(3): 403-425.
55	ARTIKA I M, MA’ROEF C N. Laboratory biosafety for handling emerging viruses[J]. Asian Pacific Journal of Tropical Biomedicine, 2017, 7(5): 483-491.
56	WELLNER A, MCMAHON C, GILMAN M S A, et al. Rapid generation of potent antibodies by autonomous hypermutation in yeast[J]. Nature Chemical Biology, 2021, 17(10): 1057-1064.
57	RIX G, WILLIAMS R L, HU V J, et al. Continuous evolution of user-defined genes at 1 million times the genomic mutation rate[J]. Science, 2024, 386(6722): eadm9073.
58	TIAN R Z, REHM F B H, CZERNECKI D, et al. Establishing a synthetic orthogonal replication system enables accelerated evolution in E. coli [J]. Science, 2024, 383(6681): 421-426.
59	MA Z Y, LI W J, SHEN Y H, et al. EvoAI enables extreme compression and reconstruction of the protein sequence space[J]. Nature Methods, 2025, 22(1): 102-112.
60	JOHNSTON K E, ALMHJELL P J, WATKINS-DULANEY E J, et al. A combinatorially complete epistatic fitness landscape in an enzyme active site[J]. Proceedings of the National Academy of Sciences of the United States of America, 2024, 121(32): e2400439121.

蛋白质语言模型 Protein language model	年份 Year	参数量 Parameters	训练数据规模 Training data size	架构 Architecture
ProtVec^[69]	2015	—	Swiss-Prot 0.55M sequences	Word2Vec
SeqVec^[70]	2019	93M	UniRef50 33M sequences	BiLSTM
UniRep^[71]	2019	18M	UniRef50 24M sequences	mLSTM
TAPE (Transformer)^[72]	2019	38M	Pfam 31M protein domains seqeunces	Encoder-only
ProGen^[73-74]	2020	1.2B	280M sequences from 5 sources	Decoder-only
ESM-1b^[75]	2021	650M	UniRef50 27.1M sequence	Encoder-only
ESM-1v^[76]	2021	650M	UniRef90 98M sequence	Encoder-only
MSA Transformer^[77]	2021	100M	26M multiple sequence alignments	Axial Transformer
ProteinBERT^[78]	2022	16M	UniRef90 106M sequences	Encoder-only
ProtGPT2^[79]	2022	738M	UniRef50 45M sequences	Decoder-only
ProtT5 (ProtTrans)^[80]	2022	3B(XL)/11B(XXL)	BFD + UniRef50 2.3B sequences	Encoder-Decoder
ESM-2^[81]	2023	8M～15B	UniRef50 60M sequences	Encoder-only
ProGen2^[82]	2023	151M～6.4B	UniRef90+BFD30 1B sequences	Decoder-only
Ankh^[83]	2023	450M(base);1.15B(large)	UniRef50 45M sequences	Encoder-Decoder
PoET^[84]	2023	604M	UniRef50 29M sets of homologous sequence	Decoder-only
ESM3^[85]	2024	1.4B～98B	3.15B protein sequences, 236M protein structures, and 539M proteins’ function annotations	Encoder-only
CARP^[86]	2024	600k～640M	UniRef50 42M sequences	CNN
ProLLaMA^[87]	2024	7B	UniRef50 48M sequences	Decoder-only
xTrimoPGLM^[88]	2024	1B～100B	UniRef90 + ColabFoldDB 939M protein sequences	Transformer
Prot42^[89]	2025	500M(base);1.1B(large)	UniRef50 57M sequences	Decoder-only
T5ProtChem^[90]	2025	102M	UniRef50 52M protein sequences and PubChem 97M SMILES sequences	Encoder-Decoder
LC-PLM^[91]	2025	1.4B	UniRef50 63M sequences	BiMamba

蛋白质语言模型 Protein language model	年份 Year	参数量 Parameters	训练数据规模 Training data size	架构 Architecture
ProtVec^[69]	2015	—	Swiss-Prot 0.55M sequences	Word2Vec
SeqVec^[70]	2019	93M	UniRef50 33M sequences	BiLSTM
UniRep^[71]	2019	18M	UniRef50 24M sequences	mLSTM
TAPE (Transformer)^[72]	2019	38M	Pfam 31M protein domains seqeunces	Encoder-only
ProGen^[73-74]	2020	1.2B	280M sequences from 5 sources	Decoder-only
ESM-1b^[75]	2021	650M	UniRef50 27.1M sequence	Encoder-only
ESM-1v^[76]	2021	650M	UniRef90 98M sequence	Encoder-only
MSA Transformer^[77]	2021	100M	26M multiple sequence alignments	Axial Transformer
ProteinBERT^[78]	2022	16M	UniRef90 106M sequences	Encoder-only
ProtGPT2^[79]	2022	738M	UniRef50 45M sequences	Decoder-only
ProtT5 (ProtTrans)^[80]	2022	3B(XL)/11B(XXL)	BFD + UniRef50 2.3B sequences	Encoder-Decoder
ESM-2^[81]	2023	8M～15B	UniRef50 60M sequences	Encoder-only
ProGen2^[82]	2023	151M～6.4B	UniRef90+BFD30 1B sequences	Decoder-only
Ankh^[83]	2023	450M(base);1.15B(large)	UniRef50 45M sequences	Encoder-Decoder
PoET^[84]	2023	604M	UniRef50 29M sets of homologous sequence	Decoder-only
ESM3^[85]	2024	1.4B～98B	3.15B protein sequences, 236M protein structures, and 539M proteins’ function annotations	Encoder-only
CARP^[86]	2024	600k～640M	UniRef50 42M sequences	CNN
ProLLaMA^[87]	2024	7B	UniRef50 48M sequences	Decoder-only
xTrimoPGLM^[88]	2024	1B～100B	UniRef90 + ColabFoldDB 939M protein sequences	Transformer
Prot42^[89]	2025	500M(base);1.1B(large)	UniRef50 57M sequences	Decoder-only
T5ProtChem^[90]	2025	102M	UniRef50 52M protein sequences and PubChem 97M SMILES sequences	Encoder-Decoder
LC-PLM^[91]	2025	1.4B	UniRef50 63M sequences	BiMamba

[1]	吴柯, 罗家豪, 李斐然. 机器学习驱动的基因组规模代谢模型构建与优化[J]. 合成生物学, 2025, 6(3): 566-584.
[2]	田晓军, 张日新. 合成基因回路面临的细胞“经济学窘境”[J]. 合成生物学, 2025, 6(3): 532-546.
[3]	章益蜻, 刘高雯. 合成生物学视角下的基因功能探索与酵母工程菌株文库构建[J]. 合成生物学, 2025, 6(3): 685-700.
[4]	黄怡, 司同, 陆安静. 生物制造标准体系建设的现状、问题与建议[J]. 合成生物学, 2025, 6(3): 701-714.
[5]	张梦瑶, 蔡鹏, 周雍进. 合成生物学助力萜类香精香料可持续生产[J]. 合成生物学, 2025, 6(2): 334-356.
[6]	张璐鸥, 徐丽, 胡晓旭, 杨滢. 合成生物学助力化妆品走进生物制造新时代[J]. 合成生物学, 2025, 6(2): 479-491.
[7]	伊进行, 唐宇琳, 李春雨, 吴鹤云, 马倩, 谢希贤. 氨基酸衍生物在化妆品中的应用及其生物合成研究进展[J]. 合成生物学, 2025, 6(2): 254-289.
[8]	韦灵珍, 王佳, 孙新晓, 袁其朋, 申晓林. 黄酮类化合物生物合成及其在化妆品中应用的研究[J]. 合成生物学, 2025, 6(2): 373-390.
[9]	肖森, 胡立涛, 石智诚, 王发银, 余思婷, 堵国成, 陈坚, 康振. 可控分子量透明质酸的生物合成研究进展[J]. 合成生物学, 2025, 6(2): 445-460.
[10]	王倩, 果士婷, 辛波, 钟成, 王钰. L-精氨酸的微生物合成研究进展[J]. 合成生物学, 2025, 6(2): 290-305.
[11]	左一萌, 张姣姣, 连佳长. 酿酒酵母使能技术在化妆品原料合成中的应用[J]. 合成生物学, 2025, 6(2): 233-253.
[12]	汤传根, 王璟, 张烁, 张昊宁, 康振. 功能肽合成和挖掘策略研究进展[J]. 合成生物学, 2025, 6(2): 461-478.
[13]	郭婷婷, 韩湘凝, 黄熙婷, 张婷婷, 孔健. 乳酸菌的合成生物学工具及在合成益肤因子中的应用[J]. 合成生物学, 2025, 6(2): 320-333.
[14]	张萍, 张维娇, 胥睿睿, 李江华, 陈坚, 康振. 防晒化合物类菌孢素氨基酸的生物合成[J]. 合成生物学, 2025, 6(2): 306-319.
[15]	黄姝涵, 马赫, 罗云孜. 生物合成红景天苷的研究进展[J]. 合成生物学, 2025, 6(2): 391-407.

AI+定向进化赋能蛋白改造及优化

AI-enabled directed evolution for protein engineering and optimization

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 4

参考文献 60

相关文章 15

编辑推荐

Metrics

本文评价