机器学习驱动的基因组规模代谢模型构建与优化

doi:10.12211/2096-8280.2024-090

合成生物学 ›› 2025, Vol. 6 ›› Issue (3): 566-584.DOI: 10.12211/2096-8280.2024-090

机器学习驱动的基因组规模代谢模型构建与优化

吴柯¹^,², 罗家豪¹^,², 李斐然¹^,²

^1.清华大学深圳国际研究生院，生物医药与健康工程研究院，广东深圳 518055
^2.清华大学化学工程系，工业生物催化教育部重点实验室，北京 100084

收稿日期:2024-12-02 修回日期:2025-02-12 出版日期:2025-06-30 发布日期:2025-06-27
通讯作者: 李斐然
作者简介:吴柯（2000—），男，博士研究生。研究方向为机器学习辅助基因组规模代谢模型开发。 E-mail：wk37@tju.edu.cn
李斐然（1993—），女，助理教授，博士生导师。研究方向为基因组规模代谢模型开发、微生物细胞工厂设计、酶参数预测，致力于构建数字生命模型。 E-mail：feiranli@sz.tsinghua.edu.cn
基金资助:
国家自然科学基金面上项目(22478223)

Applications of machine learning in the reconstruction and curation of genome-scale metabolic models

WU Ke¹^,², LUO Jiahao¹^,², LI Feiran¹^,²

^1.Institute of Biopharmaceutical and Health Engineering，Tsinghua Shenzhen International Graduate School，Tsinghua University，Shenzhen 518055，Guangdong，China
^2.Key Laboratory for Industrial Biocatalysis，Ministry of Education，Department of Chemical Engineering，Tsinghua University，Beijing 100084，China

Received:2024-12-02 Revised:2025-02-12 Online:2025-06-30 Published:2025-06-27
Contact: LI Feiran

摘要/Abstract

摘要：

自1999年首个基因组规模代谢模型（genome-scale metabolic model，GEM）问世以来，GEM已成为解析生物代谢的重要工具。该模型包含代谢基因、代谢物和反应，并结合化学计量矩阵与约束优化，系统描述和模拟生物体内的代谢过程。此外，GEM能够整合热力学参数、动力学参数、多组学数据及多细胞过程，构建更精细且具有更强大预测能力的多约束多过程模型。然而，先验知识的局限成为其发展的瓶颈。机器学习技术凭借强大的数据处理和模式识别能力，为进一步扩展GEM提供了新思路。本综述系统总结了传统GEM及多约束多过程模型的构建流程，并着重探讨了机器学习在其中关键步骤中的应用前景，如基因功能注释、途径解析、空缺填补和生物学参数预测。机器学习技术作为新的驱动力，有望大幅度提升GEM的规模和质量，深化对生物代谢机制的理解，并推动实现数字孪生细胞。

关键词: 基因组规模代谢模型, 机器学习, 合成生物学, 代谢建模, 多约束多过程模型

Abstract:

Since the publication of the first genome-scale metabolic model (GEM) in 1999, GEMs have become an essential tool for analyzing metabolism. The models integrate genes, metabolites, and reactions for combining stoichiometric matrices with constraint-based optimization to systematically describe and simulate metabolic processes in organisms. The development of automated pipelines for reconstructing GEMs has expanded their applicability to organisms from all kingdoms of life. Additionally, GEMs can integrate kinetic parameters, thermodynamic parameters, multi-omics data and multi-cellular processes to reconstruct more accurate models, thereby improving prediction accuracy. However, the reconstruction of GEMs remains heavily dependent on pre-existing knowledge, inherently limiting their scope to currently available information. This dependency restricts our ability to fully unravel the complexity and dynamic nature of metabolism. Recent advances in machine learning have demonstrated extraordinary capabilities for biological tasks such as protein structure prediction, disease identification and GEM reconstruction with functional annotation and large-scale data integration, showcasing its power in identifying patterns and uncovering hidden relationships within biological systems. Machine learning provides a promising pathway to overcome the limitations of GEMs by expanding their applicability to areas previously constrained by data availability and complexity. This review summarizes the traditional reconstruction methods of GEMs and their applications in integrating multi-dimensional data to build multi-constraint and multi-process models. The review also focuses on key applications of machine learning in gene function annotation, pathway analysis, gap-filling prediction in the reconstruction of GEMs. Additionally, the potential of machine learning in predicting kinetic, thermodynamic, and other key biochemical parameters in the reconstruction of multi-constraint and multi-process models is discussed. By combining GEMs with machine learning innovations, researchers can improve model accuracy, enhance scalability, and gain new insights into previously elusive metabolic mechanisms, bridging gaps in metabolic knowledge, and underscoring its importance as a cornerstone for future development in systems biology and biotechnology.

Key words: genome-scale metabolic model, machine learning, synthetic biology, metabolic modeling, multi-constraint and multi-process model

中图分类号:

Q814.9

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.

WU Ke, LUO Jiahao, LI Feiran. Applications of machine learning in the reconstruction and curation of genome-scale metabolic models[J]. Synthetic Biology Journal, 2025, 6(3): 566-584.

图/表 5

参考文献 147

1	GONG Z J, CHEN J Y, JIAO X Y, et al. Genome-scale metabolic network models for industrial microorganisms metabolic engineering: Current advances and future prospects[J]. Biotechnology Advances, 2024, 72: 108319.
2	ORTH J D, THIELE I, PALSSON B Ø. What is flux balance analysis?[J]. Nature Biotechnology, 2010, 28(3): 245-248.
3	WAGNER A, WANG C, FESSLER J, et al. Metabolic modeling of single Th17 cells reveals regulators of autoimmunity[J]. Cell, 2021, 184(16): 4168-4185. e21.
4	KIM B J, KIM W J, KIM D I, et al. Applications of genome-scale metabolic network model in metabolic engineering[J]. Journal of Industrial Microbiology & Biotechnology, 2015, 42(3): 339-348.
5	RYU J Y, KIM H U, LEE S Y. Reconstruction of genome-scale human metabolic models using omics data[J]. Integrative Biology, 2015, 7(8): 859-868.
6	GU C D, KIM G B, KIM W J, et al. Current status and applications of genome-scale metabolic models[J]. Genome Biology, 2019, 20(1): 121.
7	EDWARDS J S, PALSSON B O. Systems properties of the Haemophilus influenzae Rd metabolic genotype[J]. Journal of Biological Chemistry, 1999, 274(25): 17410-17416.
8	ABRAMSON J, ADLER J, DUNGER J, et al. Accurate structure prediction of biomolecular interactions with AlphaFold3[J]. Nature, 2024, 630(8016): 493-500.
9	DAUPARAS J, ANISHCHENKO I, BENNETT N, et al. Robust deep learning-based protein sequence design using ProteinMPNN[J]. Science, 2022, 378(6615): 49-56.
10	JUMPER J, EVANS R, PRITZEL A, et al. Highly accurate protein structure prediction with AlphaFold[J]. Nature, 2021, 596(7873): 583-589.
11	YU T H, CUI H Y, LI J C, et al. Enzyme function prediction using contrastive learning[J]. Science, 2023, 379(6639): 1358-1363.
12	FINNIGAN W, HEPWORTH L J, FLITSCH S L, et al. RetroBioCat as a computer-aided synthesis planning tool for biocatalytic reactions and cascades[J]. Nature Catalysis, 2021, 4(2): 98-104.
13	曾涛, 巫瑞波. 数据驱动的酶反应预测与设计[J]. 合成生物学, 2023, 4(3): 535-550.
	ZENG T, WU R B. Data-driven prediction and design for enzymatic reactions[J]. Synthetic Biology Journal, 2023, 4(3): 535-550.
14	SABZEVARI M, SZEDMAK S, PENTTILÄ M, et al. Strain design optimization using reinforcement learning[J]. PLoS Computational Biology, 2022, 18(6): e1010177.
15	禹伟, 高教琪, 周雍进.一碳生物转化合成有机酸的研究进展[J]. 合成生物学, 2024, 5(5): 1169-1188.
	YU W, GAO J Q, ZHOU Y J. Bioconversion of one carbon feedstocks for producing organic acids[J]. Synthetic Biology Journal, 2024, 5(5): 1169-1188.
16	KUNDU P, BEURA S, MONDAL S, et al. Machine learning for the advancement of genome-scale metabolic modeling[J]. Biotechnology Advances, 2024, 74: 108400.
17	KANEHISA M, FURUMICHI M, SATO Y, et al. KEGG: biological systems database as a model of the real world[J]. Nucleic Acids Research, 2025, 53(D1): D672-D677.
18	The UniProt Consortium. UniProt: the universal protein knowledgebase[J]. Nucleic Acids Research, 2018, 46(5): 2699.
19	CHANG A, JESKE L, ULBRICH S, et al. BRENDA, the ELIXIR core data resource in 2021: new developments and updates[J]. Nucleic Acids Research, 2021, 49(D1): D498-D508.
20	DUVAUD S, GABELLA C, LISACEK F, et al. Expasy, the Swiss bioinformatics resource portal, as designed by its users[J]. Nucleic Acids Research, 2021, 49(W1): W216-W227.
21	WITTIG U, REY M, WEIDEMANN A, et al. SABIO-RK: an updated resource for manually curated biochemical reaction kinetics[J]. Nucleic Acids Research, 2018, 46(D1): D656-D660.
22	BANSAL P, MORGAT A, AXELSEN K B, et al. Rhea, the reaction knowledgebase in 2022[J]. Nucleic Acids Research, 2022, 50(D1): D693-D700.
23	KARP P D, BILLINGTON R, CASPI R, et al. The BioCyc collection of microbial genomes and metabolic pathways[J]. Briefings in Bioinformatics, 2019, 20(4): 1085-1093.
24	MORETTI S, TRAN V D T, MEHL F, et al. MetaNetX/MNXref: unified namespace for metabolites and biochemical reactions in the context of metabolic models[J]. Nucleic Acids Research, 2021, 49(D1): D570-D574.
25	KING Z A, LU J, DRÄGER A, et al. BiGG Models: a platform for integrating, standardizing and sharing genome-scale models[J]. Nucleic Acids Research, 2016, 44(D1): D515-D522.
26	MALIK-SHERIFF R S, GLONT M, NGUYEN T V N, et al. BioModels-15 years of sharing computational models in life science[J]. Nucleic Acids Research, 2020, 48(D1): D407-D415.
27	ARKIN A P, COTTINGHAM R W, HENRY C S, et al. KBase: the United States Department of Energy systems biology knowledgebase[J]. Nature Biotechnology, 2018, 36(7): 566-569.
28	LIEVEN C, BEBER M E, OLIVIER B G, et al. Publisher Correction: MEMOTE for standardized genome-scale metabolic model testing[J]. Nature Biotechnology, 2020, 38(4): 504.
29	THIELE I, PALSSON B Ø. A protocol for generating a high-quality genome-scale metabolic reconstruction[J]. Nature Protocols, 2010, 5(1): 93-121.
30	HEIRENDT L, ARRECKX S, PFAU T, et al. Creation and analysis of biochemical constraint-based models using the COBRA Toolbox v.3.0[J]. Nature Protocols, 2019, 14(3): 639-702.
31	WANG H, MARCIŠAUSKAS S, SÁNCHEZ B J, et al. RAVEN 2.0: a versatile toolbox for metabolic network reconstruction and a case study on Streptomyces coelicolor [J]. PLoS Computational Biology, 2018, 14(10): e1006541.
32	DEVOID S, OVERBEEK R, DEJONGH M, et al. Automated genome annotation and metabolic model reconstruction in the SEED and Model SEED[J]. Methods in Molecular Biology, 2013, 985: 17-45.
33	MACHADO D, ANDREJEV S, TRAMONTANO M, et al. Fast automated reconstruction of genome-scale metabolic models for microbial species and communities[J]. Nucleic Acids Research, 2018, 46(15): 7542-7553.
34	CAPELA J, LAGOA D, RODRIGUES R, et al. Merlin, an improved framework for the reconstruction of high-quality genome-scale metabolic models[J]. Nucleic Acids Research, 2022, 50(11): 6052-6066.
35	ZIMMERMANN J, KALETA C, WASCHINA S. Gapseq: informed prediction of bacterial metabolic pathways and reconstruction of accurate metabolic models[J]. Genome Biology, 2021, 22(1): 81.
36	HEINKEN A, HERTEL J, ACHARYA G, et al. Genome-scale metabolic reconstruction of 7302 human microorganisms for personalized medicine[J]. Nature Biotechnology, 2023, 41(9): 1320-1331.
37	FANG X, LLOYD C J, PALSSON B O. Reconstructing organisms in silico: genome-scale models and their emerging applications[J]. Nature Reviews Microbiology, 2020, 18(12): 731-743.
38	LU H Z, XIAO L C, LIAO W B, et al. Cell factory design with advanced metabolic modelling empowered by artificial intelligence[J]. Metabolic Engineering, 2024, 85: 61-72.
39	CARRASCO MURIEL J, LONG C, SONNENSCHEIN N. Simultaneous application of enzyme and thermodynamic constraints to metabolic models using an updated Python implementation of GECKO[J]. Microbiology Spectrum, 2023, 11(6): e0170523.
40	BI X Y, CHENG Y, XU X H, et al. etiBsu1209: a comprehensive multiscale metabolic model for Bacillus subtilis [J]. Biotechnology and Bioengineering, 2023, 120(6): 1623-1639.
41	SCHROEDER W L, SUTHERS P F, WILLIS T C, et al. Current state, challenges, and opportunities in genome-scale resource allocation models: a mathematical perspective[J]. Metabolites, 2024, 14(7): 365.
42	HENRY C S, JANKOWSKI M D, BROADBELT L J, et al. Genome-scale thermodynamic analysis of Escherichia coli metabolism[J]. Biophysical Journal, 2006, 90(4): 1453-1461.
43	HENRY C S, BROADBELT L J, HATZIMANIKATIS V. Thermodynamics-based metabolic flux analysis[J]. Biophysical Journal, 2007, 92(5): 1792-1805.
44	DASH S, OLSON D G, JOSHUA CHAN S H, et al. Thermodynamic analysis of the pathway for ethanol production from cellobiose in Clostridium thermocellum [J]. Metabolic Engineering, 2019, 55: 161-169.
45	NOOR E, BAR-EVEN A, FLAMHOLZ A, et al. Pathway thermodynamics highlights kinetic obstacles in central metabolism[J]. PLoS Computational Biology, 2014, 10(2): e1003483.
46	HÄDICKE O, VON KAMP A, AYDOGAN T, et al. OptMDFpathway: identification of metabolic pathways with maximal thermodynamic driving force and its application for analyzing the endogenous CO₂ fixation potential of Escherichia coli [J]. PLoS Computational Biology, 2018, 14(9): e1006492.
47	FLAMHOLZ A, NOOR E, BAR-EVEN A, et al. Glycolytic strategy as a tradeoff between energy yield and protein cost[J]. Proceedings of the National Academy of Sciences of the United States of America, 2013, 110(24): 10039-10044.
48	BEKIARIS P S, KLAMT S. Automatic construction of metabolic models with enzyme constraints[J]. BMC Bioinformatics, 2020, 21(1): 19.
49	BEG Q K, VAZQUEZ A, ERNST J, et al. Intracellular crowding defines the mode and sequence of substrate uptake by Escherichia coli and constrains its metabolic activity[J]. Proceedings of the National Academy of Sciences of the United States of America, 2007, 104(31): 12663-12668.
50	ADADI R, VOLKMER B, MILO R, et al. Prediction of microbial growth rate versus biomass yield by a metabolic network with kinetic parameters[J]. PLoS Computational Biology, 2012, 8(7): e1002575.
51	MAO Z T, ZHAO X, YANG X, et al. ECMpy, a simplified workflow for constructing enzymatic constrained metabolic network model[J]. Biomolecules, 2022, 12(1): 65.
52	MAO Z T, NIU J H, ZHAO J X, et al. ECMpy 2.0: a Python package for automated construction and analysis of enzyme-constrained models[J]. Synthetic and Systems Biotechnology, 2024, 9(3): 494-502.
53	SÁNCHEZ B J, ZHANG C, NILSSON A, et al. Improving the phenotype predictions of a yeast genome-scale metabolic model by incorporating enzymatic constraints[J]. Molecular Systems Biology, 2017, 13(8): 935.
54	DOMENZAIN I, SÁNCHEZ B, ANTON M, et al. Reconstruction of a catalogue of genome-scale metabolic models with enzymatic constraints using GECKO 2.0[J]. Nature Communications, 2022, 13(1): 3766.
55	CHEN Y, GUSTAFSSON J, TAFUR RANGEL A, et al. Reconstruction, simulation and analysis of enzyme-constrained metabolic models using GECKO Toolbox 3.0[J]. Nature Protocols, 2024, 19(3): 629-667.
56	KERKHOVEN E J. Advances in constraint-based models: methods for improved predictive power based on resource allocation constraints[J]. Current Opinion in Microbiology, 2022, 68: 102168.
57	YANG X, MAO Z T, ZHAO X, et al. Integrating thermodynamic and enzymatic constraints into genome-scale metabolic models[J]. Metabolic Engineering, 2021, 67: 133-144.
58	LI G, HU Y T, ZRIMEC J, et al. Bayesian genome scale modelling identifies thermal determinants of yeast metabolism[J]. Nature Communications, 2021, 12(1): 190.
59	O’BRIEN E J, LERMAN J A, CHANG R L, et al. Genome-scale models of metabolism and gene expression extend and refine growth phenotype prediction[J]. Molecular Systems Biology, 2013, 9: 693.
60	ALSIYABI A, CHOWDHURY N B, LONG D N, et al. Enhancing in silico strain design predictions through next generation metabolic modeling approaches[J]. Biotechnology Advances, 2022, 54: 107806.
61	CHEN K, GAO Y, MIH N, et al. Thermosensitivity of growth is determined by chaperone-mediated proteome reallocation[J]. Proceedings of the National Academy of Sciences of the United States of America, 2017, 114(43): 11548-11553.
62	DU B, YANG L, LLOYD C J, et al. Genome-scale model of metabolism and gene expression provides a multi-scale description of acid stress responses in Escherichia coli [J]. PLoS Computational Biology, 2019, 15(12): e1007525.
63	YANG L, MIH N, ANAND A, et al. Cellular responses to reactive oxygen species are predicted from molecular mechanisms[J]. Proceedings of the National Academy of Sciences of the United States of America, 2019, 116(28): 14368-14373.
64	ZHAO J, CHEN K, PALSSON B O, et al. StressME: unified computing framework of Escherichia coli metabolism, gene expression, and stress responses[J]. PLoS Computational Biology, 2024, 20(2): e1011865.
65	CHEN Y, VAN PELT-KLEINJAN E, VAN OLST B, et al. Proteome constraints reveal targets for improving microbial fitness in nutrient-rich environments[J]. Molecular Systems Biology, 2021, 17(4): e10093.
66	ELSEMMAN I E, RODRIGUEZ PRADO A, GRIGAITIS P, et al. Whole-cell modeling in yeast predicts compartment-specific proteome constraints that drive metabolic strategies[J]. Nature Communications, 2022, 13(1): 801.
67	LI F R, CHEN Y, QI Q, et al. Improving recombinant protein production by yeast through genome-scale modeling using proteome constraints[J]. Nature Communications, 2022, 13(1): 2969.
68	CHEN Y, LI F R, MAO J W, et al. Yeast optimizes metal utilization based on metabolic network and enzyme kinetics[J]. Proceedings of the National Academy of Sciences of the United States of America, 2021, 118(12): e2020154118.
69	SALVY P, HATZIMANIKATIS V. The ETFL formulation allows multi-omics integration in thermodynamics-compliant metabolism and expression models[J]. Nature Communications, 2020, 11(1): 30.
70	OFTADEH O, SALVY P, MASID M, et al. A genome-scale metabolic model of Saccharomyces cerevisiae that integrates expression constraints and reaction thermodynamics[J]. Nature Communications, 2021, 12(1): 4790.
71	RYU J Y, KIM H U, LEE S Y. Deep learning enables high-quality and high-throughput prediction of enzyme commission numbers[J]. Proceedings of the National Academy of Sciences of the United States of America, 2019, 116(28): 13996-14001.
72	KIM G B, KIM J Y, LEE J A, et al. Functional annotation of enzyme-encoding genes using deep learning with transformer layers[J]. Nature Communications, 2023, 14(1): 7370.
73	SHI Z K, DENG R, YUAN Q Q, et al. Enzyme commission number prediction and benchmarking with hierarchical dual-core multitask learning framework[J]. Research, 2023, 6: 0153.
74	HAN S R, PARK M, KOSARAJU S, et al. Evidential deep learning for trustworthy prediction of enzyme commission number[J]. Briefings in Bioinformatics, 2023, 25(1): bbad401.
75	BUTON N, COSTE F, LE CUNFF Y. Predicting enzymatic function of protein sequences with attention[J]. Bioinformatics, 2023, 39(10): btad620.
76	WATANABE N, YAMAMOTO M, MURATA M, et al. EnzymeNet: residual neural networks model for Enzyme Commission number prediction[J]. Bioinformatics Advances, 2023, 3(1): vbad173.
77	SONG Y D, YUAN Q M, CHEN S, et al. Accurately predicting enzyme functions through geometric graph learning on ESMFold-predicted structures[J]. Nature Communications, 2024, 15(1): 8180.
78	KOCH M, DUIGOU T, FAULON J L. Reinforcement learning for bioretrosynthesis[J]. ACS Synthetic Biology, 2020, 9(1): 157-168.
79	COLEY C W, THOMAS D A, LUMMISS J A M, et al. A robotic platform for flow synthesis of organic compounds informed by AI planning[J]. Science, 2019, 365(6453): eaax1566.
80	LEVIN I, LIU M J, VOIGT C A, et al. Merging enzymatic and synthetic chemistry with computational synthesis planning[J]. Nature Communications, 2022, 13(1): 7747.
81	KREUTTER D, SCHWALLER P, REYMOND J L. Predicting enzymatic reactions with a molecular transformer[J]. Chemical Science, 2021, 12(25): 8648-8659.
82	PROBST D, MANICA M, NANA TEUKAM Y G, et al. Biocatalysed synthesis planning using data-driven learning[J]. Nature Communications, 2022, 13(1): 964.
83	ZHENG S J, ZENG T, LI C T, et al. Deep learning driven biosynthetic pathways navigation for natural products with BioNavi-NP[J]. Nature Communications, 2022, 13(1): 3342.
84	ZENG T, JIN Z H, ZHENG S J, et al. Developing BioNavi for hybrid retrosynthesis planning[J]. JACS Au, 2024, 4(7): 2492-2502.
85	KROLL A, RANJAN S, ENGQVIST M K M, et al. A general model to predict small molecule substrates of enzymes based on machine and deep learning[J]. Nature Communications, 2023, 14(1): 2787.
86	UPADHYAY V, BOORLA V S, MARANAS C D. Rank-ordering of known enzymes as starting points for re-engineering novel substrate activity using a convolutional neural network[J]. Metabolic Engineering, 2023, 78: 171-182.
87	ZHANG D C, XING H D, LIU D L, et al. Discovery of toxin-degrading enzymes with positive unlabeled deep learning[J]. ACS Catalysis, 2024, 14(5): 3336-3348.
88	SHI Z K, WANG D H, LI Y, et al. REME: an integrated platform for reaction enzyme mining and evaluation[J]. Nucleic Acids Research, 2024, 52(W1): W299-W305.
89	XING H D, CAI P L, LIU D L, et al. High-throughput prediction of enzyme promiscuity based on substrate-product pairs[J]. Briefings in Bioinformatics, 2024, 25(2): bbae089.
90	OYETUNDE T, ZHANG M H, CHEN Y X, et al. BoostGAPFILL: improving the fidelity of metabolic network reconstructions through integrated constraint and pattern-based methods[J]. Bioinformatics, 2017, 33(4): 608-611.
91	CHEN C, LIAO C, LIU Y Y. Teasing out missing reactions in genome-scale metabolic networks through hypergraph learning[J]. Nature Communications, 2023, 14(1): 2375.
92	HUANG W H, YANG F, ZHANG Q, et al. A dual-scale fused hypergraph convolution-based hyperedge prediction model for predicting missing reactions in genome-scale metabolic networks[J]. Briefings in Bioinformatics, 2024, 25(5): bbae383.
93	BOER M D, MELKONIAN C, ZAFEIROPOULOS H, et al. Improving genome-scale metabolic models of incomplete genomes with deep learning[J]. iScience, 2024, 27(12): 111349.
94	ASHBURNER M, BALL C A, BLAKE J A, et al. Gene ontology: tool for the unification of biology. The gene ontology consortium[J]. Nature Genetics, 2000, 25(1): 25-29.
95	SHEN H B, CHOU K C. EzyPred: a top-down approach for predicting enzyme functional classes and subclasses[J]. Biochemical and Biophysical Research Communications, 2007, 364(1): 53-59.
96	LI Y H, XU J Y, TAO L, et al. SVM-prot 2016: a web-server for machine learning prediction of protein functional families from sequence irrespective of similarity[J]. PLoS One, 2016, 11(8): e0155290.
97	LI Y, WANG S, UMAROV R, et al. DEEPre: sequence-based enzyme EC number prediction by deep learning[J]. Bioinformatics, 2018, 34(5): 760-769.
98	NURSIMULU N, XU L L, WASMUTH J D, et al. Improved enzyme annotation with EC-specific cutoffs using DETECT v2[J]. Bioinformatics, 2018, 34(19): 3393-3395.
99	DALKIRAN A, RIFAIOGLU A S, MARTIN M J, et al. ECPred: a tool for the prediction of the enzymatic functions of protein sequences based on the EC nomenclature[J]. BMC Bioinformatics, 2018, 19(1): 334.
100	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.
101	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.
102	HADADI N, HAFNER J, SHAJKOFCI A, et al. ATLAS of biochemistry: a repository of all possible biochemical reactions for synthetic biology and metabolic engineering studies[J]. ACS Synthetic Biology, 2016, 5(10): 1155-1166.
103	KHERSONSKY O, TAWFIK D S. Enzyme promiscuity: a mechanistic and evolutionary perspective[J]. Annual Review of Biochemistry, 2010, 79: 471-505.
104	PONTRELLI S, FRICKE R C B, TEOH S T, et al. Metabolic repair through emergence of new pathways in Escherichia coli [J]. Nature Chemical Biology, 2018, 14(11): 1005-1009.
105	NI Z F, STINE A E, TYO K E J, et al. Curating a comprehensive set of enzymatic reaction rules for efficient novel biosynthetic pathway design[J]. Metabolic Engineering, 2021, 65: 79-87.
106	NAM H, LEWIS N E, LERMAN J A, et al. Network context and selection in the evolution to enzyme specificity[J]. Science, 2012, 337(6098): 1101-1104.
107	MORIYA Y, SHIGEMIZU D, HATTORI M, et al. PathPred: an enzyme-catalyzed metabolic pathway prediction server[J]. Nucleic Acids Research, 2010, 38(S2): W138-W143.
108	DELÉPINE B, DUIGOU T, CARBONELL P, et al. RetroPath2.0: a retrosynthesis workflow for metabolic engineers[J]. Metabolic Engineering, 2018, 45: 158-170.
109	DING S Z, TIAN Y, CAI P L, et al. novoPathFinder: a webserver of designing novel-pathway with integrating GEM-model[J]. Nucleic Acids Research, 2020, 48(W1): W477-W487.
110	GENHEDEN S, THAKKAR A, CHADIMOVÁ V, et al. AiZynthFinder: a fast, robust and flexible open-source software for retrosynthetic planning[J]. Journal of Cheminformatics, 2020, 12(1): 70.
111	GRICOURT G, MEYER P, DUIGOU T, et al. Artificial intelligence methods and models for retro-biosynthesis: a scoping review[J]. ACS Synthetic Biology, 2024, 13(8): 2276-2294.
112	COLEY C W, ROGERS L, GREEN W H, et al. SCScore: synthetic complexity learned from a reaction corpus[J]. Journal of Chemical Information and Modeling, 2018, 58(2): 252-261.
113	CARBONELL P, WONG J, SWAINSTON N, et al. Selenzyme: enzyme selection tool for pathway design[J]. Bioinformatics, 2018, 34(12): 2153-2154.
114	RAHMAN S A, CUESTA S M, FURNHAM N, et al. EC-BLAST: a tool to automatically search and compare enzyme reactions[J]. Nature Methods, 2014, 11(2): 171-174.
115	GIRI V, SIVAKUMAR T V, CHO K M, et al. RxnSim: a tool to compare biochemical reactions[J]. Bioinformatics, 2015, 31(22): 3712-3714.
116	STONEY R A, HANKO E K R, CARBONELL P, et al. SelenzymeRF: updated enzyme suggestion software for unbalanced biochemical reactions[J]. Computational and Structural Biotechnology Journal, 2023, 21: 5868-5876.
117	WILLETT P. Searching techniques for databases of two- and three-dimensional chemical structures[J]. Journal of Medicinal Chemistry, 2005, 48(13): 4183-4199.
118	SCHWALLER P, HOOVER B, REYMOND J L, et al. Extraction of organic chemistry grammar from unsupervised learning of chemical reactions[J]. Science Advances, 2021, 7(15): eabe4166.
119	QIAN W J, WANG X R, HUANG Y S, et al. Deep learning-driven insights into enzyme-substrate interaction discovery[J]. Journal of Chemical Information and Modeling, 2025, 65(1): 187-200.
120	ORTH J D, PALSSON B Ø. Systematizing the generation of missing metabolic knowledge[J]. Biotechnology and Bioengineering, 2010, 107(3): 403-412.
121	THIELE I, VLASSIS N, FLEMING R M T. fastGapFill: efficient gap filling in metabolic networks[J]. Bioinformatics, 2014, 30(17): 2529-2531.
122	KUMAR V S, DASIKA M S, MARANAS C D. Optimization based automated curation of metabolic reconstructions[J]. BMC Bioinformatics, 2007, 8: 212.
123	HECKMANN D, LLOYD C J, MIH N, et al. Machine learning applied to enzyme turnover numbers reveals protein structural correlates and improves metabolic models[J]. Nature Communications, 2018, 9(1): 5252.
124	LI F R, YUAN L, LU H Z, et al. Deep learning-based k_cat prediction enables improved enzyme-constrained model reconstruction[J]. Nature Catalysis, 2022, 5(8): 662-672.
125	KROLL A, ROUSSET Y, HU X P, et al. Turnover number predictions for kinetically uncharacterized enzymes using machine and deep learning[J]. Nature Communications, 2023, 14(1): 4139.
126	QIU S Z, ZHAO S M, YANG A D. DLTKcat: deep learning-based prediction of temperature-dependent enzyme turnover rates[J]. Briefings in Bioinformatics, 2023, 25(1): bbad506.
127	WANG T, XIANG G M, HE S W, et al. DeepEnzyme: a robust deep learning model for improved enzyme turnover number prediction by utilizing features of protein 3D-structures[J]. Briefings in Bioinformatics, 2024, 25(5): bbae409.
128	KROLL A, ENGQVIST M K M, HECKMANN D, et al. Deep learning allows genome-scale prediction of Michaelis constants from structural features[J]. PLoS Biology, 2021, 19(10): e3001402.
129	HE X, YAN M. GraphKM: machine and deep learning for K_m prediction of wildtype and mutant enzymes[J]. BMC Bioinformatics, 2024, 25(1): 135.
130	MAEDA K, HATAE A, SAKAI Y, et al. MLAGO: machine learning-aided global optimization for Michaelis constant estimation of kinetic modeling[J]. BMC Bioinformatics, 2022, 23(1): 455.
131	WANG J J, YANG Z J, CHEN C, et al. MPEK: a multitask deep learning framework based on pretrained language models for enzymatic reaction kinetic parameters prediction[J]. Briefings in Bioinformatics, 2024, 25(5): bbae387.
132	YU H, DENG H X, HE J H, et al. UniKP: a unified framework for the prediction of enzyme kinetic parameters[J]. Nature Communications, 2023, 14(1): 8211.
133	SHEN X W, CUI Z H, LONG J Y, et al. EITLEM-Kinetics: a deep-learning framework for kinetic parameter prediction of mutant enzymes[J]. Chem Catalysis, 2024, 4(9): 101094.
134	WANG L, UPADHYAY V, MARANAS C D. dGPredictor: automated fragmentation method for metabolic reaction free energy prediction and de novo pathway design[J]. PLoS Computational Biology, 2021, 17(9): e1009448.
135	ALAZMI M, KUWAHARA H, SOUFAN O, et al. Systematic selection of chemical fingerprint features improves the Gibbs energy prediction of biochemical reactions[J]. Bioinformatics, 2019, 35(15): 2634-2643.
136	JUNG F, FREY K, ZIMMER D, et al. DeepSTABp: a deep learning approach for the prediction of thermal protein stability[J]. International Journal of Molecular Sciences, 2023, 24(8): 7444.
137	LI M Y, WANG H Z, YANG Z W, et al. DeepTM: a deep learning algorithm for prediction of melting temperature of thermophilic proteins directly from sequences[J]. Computational and Structural Biotechnology Journal, 2023, 21: 5544-5560.
138	LI G, RABE K S, NIELSEN J, et al. Machine learning applied to predicting microorganism growth temperatures and enzyme catalytic optima[J]. ACS Synthetic Biology, 2019, 8(6): 1411-1420.
139	GADO J E, BECKHAM G T, PAYNE C M. Improving enzyme optimum temperature prediction with resampling strategies and ensemble learning[J]. Journal of Chemical Information and Modeling, 2020, 60(8): 4098-4107.
140	LI G, BURIC F, ZRIMEC J, et al. Learning deep representations of enzyme thermal adaptation[J]. Protein Science, 2022, 31(12): e4480.
141	ZHANG Y, GUAN F F, XU G S, et al. A novel thermophilic chitinase directly mined from the marine metagenome using the deep learning tool Preoptem[J]. Bioresources and Bioprocessing, 2022, 9(1): 54.
142	NILSSON A, NIELSEN J, PALSSON B O. Metabolic models of protein allocation call for the kinetome[J]. Cell Systems, 2017, 5(6): 538-541.
143	SUTHERS P F, FOSTER C J, SARKAR D, et al. Recent advances in constraint and machine learning-based metabolic modeling by leveraging stoichiometric balances, thermodynamic feasibility and kinetic law formalisms[J]. Metabolic Engineering,2021, 63: 13-33.
144	JANKOWSKI M D, HENRY C S, BROADBELT L J, et al. Group contribution method for thermodynamic analysis of complex metabolic networks[J]. Biophysical Journal, 2008, 95(3): 1487-1499.
145	NOOR E, HARALDSDÓTTIR H S, MILO R, et al. Consistent estimation of Gibbs energy using component contributions[J]. PLoS Computational Biology, 2013, 9(7): e1003098.
146	BEBER M E, GOLLUB M G, MOZAFFARI D, et al. eQuilibrator 3.0: a database solution for thermodynamic constant estimation[J]. Nucleic Acids Research, 2022, 50(D1): D603-D609.
147	LEUENBERGER P, GANSCHA S, KAHRAMAN A, et al. Cell-wide analysis of protein thermal unfolding reveals determinants of thermostability[J]. Science, 2017, 355(6327): eaai7825.

模型应用	方法	模型框架	输入	输出	特点
辅助基因注释	DeepEC^[71]	CNN	氨基酸序列	EC编号	可区分酶与非酶，无法进行多功能注释
	CLEAN^[11]	预训练蛋白质大语言模型（ESM-1b），对比学习	氨基酸序列	EC编号	无法区分酶与非酶，可进行多功能注释，可用于数据极少的EC编号
	DeepECtransformer^[72]	预训练蛋白质大语言模型（ProtBert），Transformer	氨基酸序列	EC编号	可区分酶与非酶，可进行多功能注释，不可用于数据极少的EC编号
	ECRECer^[73]	预训练蛋白质大语言模型（ESM-1b），GRU	氨基酸序列	EC编号	可区分酶与非酶，可进行多功能注释
	ECPICK^[74]	One-hot，CNN	氨基酸序列	EC编号	无法区分酶与非酶，不可进行多功能注释，可输出预测EC编号的置信度
	EnzBert^[75]	Transformer	氨基酸序列	EC编号	无法区分酶与非酶，不可进行多功能注释，可推测关键残基
	EnzymeNet^[76]	CNN，ResNet	氨基酸序列	EC编号	可区分酶与非酶，无法进行多功能注释
	GraphEC^[77]	预训练蛋白质大语言模型（ProtTrans），ESMFold	氨基酸序列和蛋白质三维结构	EC编号	无法区分酶与非酶，可进行多功能注释，计算资源需求高
辅助途径解析-反应预测	RetroPath RL^[78]	基于蒙特卡洛树搜索的强化学习方法	化合物SMILES	合成途径	缓解组合爆炸问题，允许探索深度是只基于模板版本的两倍以上
	ASKCOS^[79]	FNN	化合物SMILES	合成途径	缓解组合爆炸问题，适用于化学合成途径设计
	chemoenzymatic-ASKCOS^[80]	DNN	化合物SMILES	合成途径	缓解组合爆炸问题，可进行化学合成与生物合成混合途径设计
	RetroBioCat^[12]	DNN	化合物SMILES	合成途径	缓解组合爆炸问题, 处理大分子化合物存在挑战
	Kreutter等开发的方法^[81]	Transformer	化合物SMILES和酶功能描述	反应产物	预测单步反应，无法给出完整的反应，无法推荐酶功能
	Probst等^[82]	Transformer	化合物SMILES和EC编号	反应产物	预测单步反应，无法给出完整的反应，无法推荐EC编号
	BioNavi-NP^[83]	Transformer	天然产SMILES	合成途径	针对天然产物及类似物，可预测多步途径
	BioNavi^[84]	Transformer	化合物SMILES	合成途径	可进行化学合成与生物合成混合途径设计
辅助途径解析-酶挖掘	ESP^[85]	预训练蛋白质大语言模型（ESM-1b），GNN，XGBoost	氨基酸序列和分子指纹	酶-底物结合可能性	对于训练集中没出现代谢物的预测性能会有明显下降
	EnzRank^[86]	分子指纹，CNN	氨基酸序列和分子指纹	酶-底物结合可能性	对于天然底物及其相似物具有良好的区分能力
	PU-EPP^[87]	GNN，正样本和无标签学习	氨基酸序列和化合物SMILES	酶-底物结合可能性	鲁棒性强，可鉴定酶和底物的关键位点
	MEI	预训练蛋白质大语言模型（ESM-1b），GNN，DNN	氨基酸序列和化合物SMILES	酶-底物结合可能性	可利用专业数据集微调进行特定任务预测
	REME^[88]	集成ESP、DLKcat/TurNuP、DeepET等模型	反应SMILES	酶列表	多维度评价与筛选有效提高了推荐酶列表的可信度
	SPEPP^[89]	Word2Vec，Transformer	底物、反应物和酶	酶催化底物-产物反应的可能性	计算效率相较基于相似性的方法显著提高，可大规模使用，需要提供候选酶集合
辅助空缺填补	BoostGAPFILL^[90]	基于矩阵分解的推荐系统技术	代谢模型	填补反应	融合了拓扑和约束方法，能够识别代谢网络中的潜在模式
	CHESHIRE^[91]	GCN	代谢模型	填补反应	计算效率高，可解释性强，可能引入假阳性反应，缺少反应方向性信息
	DSHCNet^[92]	GCN，MLP	代谢模型	填补反应	对反应数据依赖较强，在适应不同反应数据库中存在挑战
	DNNGIOR^[93]	CNN	代谢模型	填补反应	预测性能受系统发育距离影响

模型应用	方法	模型框架	输入	输出	特点
辅助基因注释	DeepEC^[71]	CNN	氨基酸序列	EC编号	可区分酶与非酶，无法进行多功能注释
	CLEAN^[11]	预训练蛋白质大语言模型（ESM-1b），对比学习	氨基酸序列	EC编号	无法区分酶与非酶，可进行多功能注释，可用于数据极少的EC编号
	DeepECtransformer^[72]	预训练蛋白质大语言模型（ProtBert），Transformer	氨基酸序列	EC编号	可区分酶与非酶，可进行多功能注释，不可用于数据极少的EC编号
	ECRECer^[73]	预训练蛋白质大语言模型（ESM-1b），GRU	氨基酸序列	EC编号	可区分酶与非酶，可进行多功能注释
	ECPICK^[74]	One-hot，CNN	氨基酸序列	EC编号	无法区分酶与非酶，不可进行多功能注释，可输出预测EC编号的置信度
	EnzBert^[75]	Transformer	氨基酸序列	EC编号	无法区分酶与非酶，不可进行多功能注释，可推测关键残基
	EnzymeNet^[76]	CNN，ResNet	氨基酸序列	EC编号	可区分酶与非酶，无法进行多功能注释
	GraphEC^[77]	预训练蛋白质大语言模型（ProtTrans），ESMFold	氨基酸序列和蛋白质三维结构	EC编号	无法区分酶与非酶，可进行多功能注释，计算资源需求高
辅助途径解析-反应预测	RetroPath RL^[78]	基于蒙特卡洛树搜索的强化学习方法	化合物SMILES	合成途径	缓解组合爆炸问题，允许探索深度是只基于模板版本的两倍以上
	ASKCOS^[79]	FNN	化合物SMILES	合成途径	缓解组合爆炸问题，适用于化学合成途径设计
	chemoenzymatic-ASKCOS^[80]	DNN	化合物SMILES	合成途径	缓解组合爆炸问题，可进行化学合成与生物合成混合途径设计
	RetroBioCat^[12]	DNN	化合物SMILES	合成途径	缓解组合爆炸问题, 处理大分子化合物存在挑战
	Kreutter等开发的方法^[81]	Transformer	化合物SMILES和酶功能描述	反应产物	预测单步反应，无法给出完整的反应，无法推荐酶功能
	Probst等^[82]	Transformer	化合物SMILES和EC编号	反应产物	预测单步反应，无法给出完整的反应，无法推荐EC编号
	BioNavi-NP^[83]	Transformer	天然产SMILES	合成途径	针对天然产物及类似物，可预测多步途径
	BioNavi^[84]	Transformer	化合物SMILES	合成途径	可进行化学合成与生物合成混合途径设计
辅助途径解析-酶挖掘	ESP^[85]	预训练蛋白质大语言模型（ESM-1b），GNN，XGBoost	氨基酸序列和分子指纹	酶-底物结合可能性	对于训练集中没出现代谢物的预测性能会有明显下降
	EnzRank^[86]	分子指纹，CNN	氨基酸序列和分子指纹	酶-底物结合可能性	对于天然底物及其相似物具有良好的区分能力
	PU-EPP^[87]	GNN，正样本和无标签学习	氨基酸序列和化合物SMILES	酶-底物结合可能性	鲁棒性强，可鉴定酶和底物的关键位点
	MEI	预训练蛋白质大语言模型（ESM-1b），GNN，DNN	氨基酸序列和化合物SMILES	酶-底物结合可能性	可利用专业数据集微调进行特定任务预测
	REME^[88]	集成ESP、DLKcat/TurNuP、DeepET等模型	反应SMILES	酶列表	多维度评价与筛选有效提高了推荐酶列表的可信度
	SPEPP^[89]	Word2Vec，Transformer	底物、反应物和酶	酶催化底物-产物反应的可能性	计算效率相较基于相似性的方法显著提高，可大规模使用，需要提供候选酶集合
辅助空缺填补	BoostGAPFILL^[90]	基于矩阵分解的推荐系统技术	代谢模型	填补反应	融合了拓扑和约束方法，能够识别代谢网络中的潜在模式
	CHESHIRE^[91]	GCN	代谢模型	填补反应	计算效率高，可解释性强，可能引入假阳性反应，缺少反应方向性信息
	DSHCNet^[92]	GCN，MLP	代谢模型	填补反应	对反应数据依赖较强，在适应不同反应数据库中存在挑战
	DNNGIOR^[93]	CNN	代谢模型	填补反应	预测性能受系统发育距离影响

参数类型	方法	模型框架	输入	输出	特点
动力学参数	Heckmann等开发的方法^[123]	随机森林，MLP	GEM/蛋白质结构/EC号首位/pH等	k_cat	可预测体内k_cat，适用于大肠杆菌
	DLKcat^[124]	GNN，CNN	氨基酸序列和化合物SMILES	k_cat	预测k_cat（R²=0.44），内置于GECKO 3.0
	TurNuP^[125]	预训练蛋白质大语言模型（ESM-1b），XGBoost	氨基酸序列和反应指纹	k_cat	预测k_cat（R²=0.44），无法区分多底物反应中不同底物的k_cat
	DLTKcat^[126]	GNN，CNN	氨基酸序列，化合物SMILES和温度	k_cat	预测不同温度下的k_cat（R²=0.66）
	DeepEnzyme^[127]	GCN	氨基酸序列，化合物SMILES和蛋白质三维结构	k_cat	预测k_cat（R²=0.58），预测突变型需要具备蛋白质结构预测能力
	Kroll等开发的方法^[128]	预训练蛋白质大语言模型（ESM-1b）	氨基酸序列和分子指纹	K_m	预测K_m（R²=0.53）
	GraphKM^[129]	预训练蛋白质大语言模型（ESM-2），GNN	氨基酸序列和化合物SMILES	K_m	预测K_m（R²=0.62），模型的预测性能受限于训练数据集的规模和质量
	MLAGO^[130]	随机森林	EC编号，KEGG ID和物种编号	K_m	预测K_m（R²=0.53），泛化能力受限于EC编号、KEGG ID、物种编号信息
	MPEK^[131]	预训练蛋白质大语言模型（ProtT5），预训练小分子大语言模型（Mole-BERT），多任务学习	氨基酸序列和化合物SMILES	k_cat和K_m	支持同时预测k_cat（R²=0.64）和K_m（R²=0.60）
	UniKP^[132]	预训练蛋白质大语言模型（UniRef50），预训练小分子大语言模型（SMILES Transformer），极度随机树	氨基酸序列和化合物SMILES	k_cat、K_m和k_cat/K_m	支持分别预测k_cat（R²=0.67）、K_m（R²=0.60）和k_cat/K_m（R²=0.56），鲁棒性强，支持温度和pH输入
	EITLEM-Kinetics^[133]	预训练蛋白质大语言模型（ESM-1v），迁移学习	氨基酸序列和化合物SMILES	k_cat、K_m和k_cat/K_m	支持分别预测k_cat（R²=0.72），K_m（R²=0.69）和k_cat/K_m（R²=0.68），蛋白突变体的k_cat预测性能优异
热力学参数	dGPredictor^[134]	线性回归模型	分子指纹	∆_rG′^⊖	不适用于异构化反应与涉及金属或聚合物结构的反应
热力学参数	Alazmi等开发的方法^[135]	线性回归模型	分子指纹	∆_rG′^⊖	特征提取方式通用性强，可用于非天然反应
温度相关参数	DeepSTABp^[136]	预训练蛋白质大语言模型（ProtTrans），MLP	氨基酸序列、生物体生长温度和实验条件	T_m	预测能力对点突变不敏感
	DeepTM^[137]	GCN	氨基酸序列	T_m	特征提取复杂，训练数据未考虑其他对蛋白熔解温度影响的因素
	Tome^[138]	SVR，随机森林	氨基酸序列，OGT	T_opt	训练集高于85 °C的T_opt值占比不足5%，限制了Tome对高温稳定性酶的预测能力
	TOMER^[139]	集成学习	氨基酸序列，OGT	T_opt	重采样缓解了数据分布不平衡，对高于85 ℃的T_opt值预测性能显著提升
	DeepET^[140]	ResNet，迁移学习	氨基酸序列	T_m和T_opt	类似于Tome，数据分布不均衡会限制其对极端温度蛋白质的预测性能
	Preoptem^[141]	One-hot，CNN	氨基酸序列	T_opt	Pearson相关系数r=0.58，适用于嗜热蛋白

参数类型	方法	模型框架	输入	输出	特点
动力学参数	Heckmann等开发的方法^[123]	随机森林，MLP	GEM/蛋白质结构/EC号首位/pH等	k_cat	可预测体内k_cat，适用于大肠杆菌
	DLKcat^[124]	GNN，CNN	氨基酸序列和化合物SMILES	k_cat	预测k_cat（R²=0.44），内置于GECKO 3.0
	TurNuP^[125]	预训练蛋白质大语言模型（ESM-1b），XGBoost	氨基酸序列和反应指纹	k_cat	预测k_cat（R²=0.44），无法区分多底物反应中不同底物的k_cat
	DLTKcat^[126]	GNN，CNN	氨基酸序列，化合物SMILES和温度	k_cat	预测不同温度下的k_cat（R²=0.66）
	DeepEnzyme^[127]	GCN	氨基酸序列，化合物SMILES和蛋白质三维结构	k_cat	预测k_cat（R²=0.58），预测突变型需要具备蛋白质结构预测能力
	Kroll等开发的方法^[128]	预训练蛋白质大语言模型（ESM-1b）	氨基酸序列和分子指纹	K_m	预测K_m（R²=0.53）
	GraphKM^[129]	预训练蛋白质大语言模型（ESM-2），GNN	氨基酸序列和化合物SMILES	K_m	预测K_m（R²=0.62），模型的预测性能受限于训练数据集的规模和质量
	MLAGO^[130]	随机森林	EC编号，KEGG ID和物种编号	K_m	预测K_m（R²=0.53），泛化能力受限于EC编号、KEGG ID、物种编号信息
	MPEK^[131]	预训练蛋白质大语言模型（ProtT5），预训练小分子大语言模型（Mole-BERT），多任务学习	氨基酸序列和化合物SMILES	k_cat和K_m	支持同时预测k_cat（R²=0.64）和K_m（R²=0.60）
	UniKP^[132]	预训练蛋白质大语言模型（UniRef50），预训练小分子大语言模型（SMILES Transformer），极度随机树	氨基酸序列和化合物SMILES	k_cat、K_m和k_cat/K_m	支持分别预测k_cat（R²=0.67）、K_m（R²=0.60）和k_cat/K_m（R²=0.56），鲁棒性强，支持温度和pH输入
	EITLEM-Kinetics^[133]	预训练蛋白质大语言模型（ESM-1v），迁移学习	氨基酸序列和化合物SMILES	k_cat、K_m和k_cat/K_m	支持分别预测k_cat（R²=0.72），K_m（R²=0.69）和k_cat/K_m（R²=0.68），蛋白突变体的k_cat预测性能优异
热力学参数	dGPredictor^[134]	线性回归模型	分子指纹	∆_rG′^⊖	不适用于异构化反应与涉及金属或聚合物结构的反应
热力学参数	Alazmi等开发的方法^[135]	线性回归模型	分子指纹	∆_rG′^⊖	特征提取方式通用性强，可用于非天然反应
温度相关参数	DeepSTABp^[136]	预训练蛋白质大语言模型（ProtTrans），MLP	氨基酸序列、生物体生长温度和实验条件	T_m	预测能力对点突变不敏感
	DeepTM^[137]	GCN	氨基酸序列	T_m	特征提取复杂，训练数据未考虑其他对蛋白熔解温度影响的因素
	Tome^[138]	SVR，随机森林	氨基酸序列，OGT	T_opt	训练集高于85 °C的T_opt值占比不足5%，限制了Tome对高温稳定性酶的预测能力
	TOMER^[139]	集成学习	氨基酸序列，OGT	T_opt	重采样缓解了数据分布不平衡，对高于85 ℃的T_opt值预测性能显著提升
	DeepET^[140]	ResNet，迁移学习	氨基酸序列	T_m和T_opt	类似于Tome，数据分布不均衡会限制其对极端温度蛋白质的预测性能
	Preoptem^[141]	One-hot，CNN	氨基酸序列	T_opt	Pearson相关系数r=0.58，适用于嗜热蛋白

[1]	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.
[2]	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.
[3]	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.
[4]	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.
[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.

机器学习驱动的基因组规模代谢模型构建与优化

Applications of machine learning in the reconstruction and curation of genome-scale metabolic models

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 5

参考文献 147

相关文章 15

编辑推荐

Metrics

本文评价