抗体药物智能开发的深度学习策略

doi:10.19586/j.2095-2341.2025.0107

生物技术进展 ›› 2026, Vol. 16 ›› Issue (1): 29-37.DOI: 10.19586/j.2095-2341.2025.0107

抗体药物智能开发的深度学习策略

曾鑫(), 王辰, 王正俊, 梁何楚, 张正平()

正大天晴药业集团股份有限公司，南京 211122

收稿日期:2025-08-18 接受日期:2025-10-22 出版日期:2026-01-25 发布日期:2026-02-12
通讯作者: 张正平
作者简介:曾鑫 E-mail: xinzeng@cttq.com；
基金资助:
连云港市基础研究计划（青年科技创新人才项目）资助项目(JCYJ2407)

Deep Learning Strategies for Intelligent Development of Antibody Drugs

Xin ZENG(), Chen WANG, Zhengjun WANG, Hechu LIANG, Zhengping ZHANG()

Chia Tai Tianqing Pharmaceutical Group Co. ，Ltd. ，Nanjing 211122，China

Received:2025-08-18 Accepted:2025-10-22 Online:2026-01-25 Published:2026-02-12
Contact: Zhengping ZHANG

摘要/Abstract

摘要：

抗体药物开发面临周期长(>3年)、成本高(>2亿美元)以及多属性协同优化困难等产业瓶颈。传统方法如杂交瘤技术存在通量低、全局优化能力不足等局限。近年来，深度学习（deep learning，DL）技术为抗体药物的智能化开发提供了突破性解决方案。系统综述了DL在抗体药物开发中的研究进展，重点探讨了抗体序列设计、结构预测、亲和力预测与成熟、多目标优化等核心环节的代表性方法与技术挑战，并对未来发展进行了展望，以期为抗体药物研发向智能化、全局化方向转型提供参考。

关键词: 深度学习, 抗体药物, 多目标协同优化

Abstract:

The development of antibody drugs faces significant industrial challenges， including extended timelines （>3 years）， high costs （>$200 million） and difficulties in collaborative optimization of multiple attributes. Traditional methods such as hybridoma technology are limited by low throughput and inadequate global optimization capabilities. In recent years， deep learning （DL） has provided breakthrough solutions for the intelligent development of antibody drugs. This review systematically summarized the research progress of DL in antibody drug development， with a focus on exploring representative methods and technical challenges in core aspects such as antibody sequence design， structure prediction， affinity prediction and maturation， and multi-objective optimization. It also provides an outlook on future development， aiming to provide a reference for the transformation of antibody drug research and development towards intelligence and globalization.

Key words: deep learning, antibody drugs, multi-objective collaborative optimization

中图分类号:

Q816

曾鑫, 王辰, 王正俊, 梁何楚, 张正平. 抗体药物智能开发的深度学习策略[J]. 生物技术进展, 2026, 16(1): 29-37.

Xin ZENG, Chen WANG, Zhengjun WANG, Hechu LIANG, Zhengping ZHANG. Deep Learning Strategies for Intelligent Development of Antibody Drugs[J]. Current Biotechnology, 2026, 16(1): 29-37.

图/表 4

参考文献 51

[1]	WANG Z, WANG G, LU H, et al.. Development of therapeutic antibodies for the treatment of diseases[J/OL]. Mol. Biomed., 2022, 3(1): 35[2025-12-21]. .
[2]	DEWAKER V, MORYA V K, KIM Y H, et al.. Revolutionizing oncology: the role of artificial intelligence (AI) as an antibody design, and optimization tools[J/OL]. Biomark. Res., 2025, 13(1): 52[2025-12-21]. .
[3]	MATSUNAGA R, TSUMOTO K. Accelerating antibody discovery and optimization with high-throughput experimentation and machine learning[J/OL]. J. Biomed. Sci., 2025, 32(1): 46[2025-12-21]. .
[4]	纪思佳,张小雪,翟丽丽,等.抗体药物筛选技术的研究进展[J].生物技术进展,2025,15(1):43-49.
	JI S J, ZHANG X X, ZHAI L L, et al.. Research progress of antibody drug screening technology[J]. Curr. Biotechnol., 2025, 15(1): 43-49.
[5]	OLSEN T H, MOAL I H, DEANE C M. AbLang: an antibody language model for completing antibody sequences[J/OL]. Bioinform. Adv., 2022, 2: vbac046[2025-12-21]. .
[6]	RUFFOLO J A, GRAY J J, SULAM J. Deciphering antibody affinity maturation with language models and weakly supervised learning[EB/OL]. arXiv, 2021: 2112.07782[2025-09-30]. .
[7]	SHUAI R W, RUFFOLO J A, GRAY J J. IgLM: infilling language modeling for antibody sequence design[J]. Cell Syst., 2023, 14(11): 979-989.
[8]	HADSUND J T, SATŁAWA T, JANUSZ B, et al.. nanoBERT: a deep learning model for gene agnostic navigation of the nanobody mutational space[J/OL]. Bioinform. Adv., 2024, 4(1): vbae033[2025-09-30]. .
[9]	NIJKAMP E, RUFFOLO J A, WEINSTEIN E N, et al.. ProGen2: exploring the boundaries of protein language models[J]. Cell Syst., 2023, 14(11): 968-978.
[10]	MADANI A, KRAUSE B, GREENE E R, et al.. Large language models generate functional protein sequences across diverse families[J]. Nat. Biotechnol., 2023, 41(8): 1099-1106.
[11]	CAI H, ZHANG Z, WANG M, et al.. Pretrainable geometric graph neural network for antibody affinity maturation[J/OL]. Nat. Commun., 2024, 15(1): 7785[2025-12-21]. .
[12]	JIN W, WOHLWEND J, BARZILAY R, et al.. Iterative refinement graph neural network for antibody sequence-structure co-design[J/OL]. arXiv, 2021: 2110.04624[2025-09-30]. .
[13]	DAUPARAS J, ANISHCHENKO I, BENNETT N, et al.. Robust deep learning-based protein sequence design using ProteinMPNN[J]. Science, 2022, 378(6615): 49-56.
[14]	MARTINKUS K, LUDWICZAK J, CHO K, et al.. AbDiffuser: full-atom generation of in vitro functioning antibodies[J/OL]. arXiv, 2023: 2308.05027[2025-09-30]. .
[15]	WATSON J L, JUERGENS D, BENNETT N R, et al.. De novo design of protein structure and function with RFdiffusion[J]. Nature, 2023, 620(7976): 1089-1100.
[16]	INGRAHAM J B, BARANOV M, COSTELLO Z, et al.. Illuminating protein space with a programmable generative model[J]. Nature, 2023, 623(7989): 1070-1078.
[17]	LI Y, LANG Y, XU C, et al.. Benchmarking inverse folding models for antibody CDR sequence design[J/OL]. PLoS One, 2025, 20(6): e0324566[2025-12-21]. .
[18]	LIN Z, 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.
[19]	KENLAY H, DREYER F A, KOVALTSUK A, et al.. Large scale paired antibody language models[J/OL]. PLoS Comput. Biol., 2024, 20(12): e1012646[2025-12-21]. .
[20]	WEISSENOW K, ROST B. Are protein language models the new universal key?[J/OL]. Curr. Opin. Struct. Biol., 2025, 91: 102997[2025-12-21]. .
[21]	WANG L, LI X, ZHANG H, et al.. A comprehensive review of protein language models[J/OL]. arXiv, 2025: 2502.06881[2025-09-30]. .
[22]	BENNETT N R, WATSON J L, RAGOTTE R J, et al.. Atomically accurate de novo design of antibodies with RFdiffusion[J]. Nature, 2025, 649(8095): 183-193.
[23]	LUO S, SU Y, PENG X, et al.. Antigen-specific antibody design and optimization with diffusion-based generative models for protein structures[J]. Adv. Neural. Inf. Process Syst., 2022, 35: 9754-9767.
[24]	HU Y, TAO F, XU J, et al.. Combining transformer and 3DCNN models to achieve co-design of structures and sequences of antibodies in a diffusional manner[J/OL]. J. Pharm. Anal., 2025, 15(6): 101267[2025-10-21]. .
[25]	WU J, KONG X, SUN N, et al.. FlowDesign: improved design of antibody CDRs through flow matching and better prior distributions[J/OL]. Cell Syst., 2025, 16(6): 101270[2025-10-21]. .
[26]	ABANADES B, WONG W K, BOYLES F, et al.. ImmuneBuilder: deep-Learning models for predicting the structures of immune proteins[J/OL]. Commun. Biol., 2023, 6(1): 575[2025-12-21]. .
[27]	WU J, WU F, JIANG B, et al.. tFold-Ab: fast and accurate antibody structure prediction without sequence homologs[J/OL]. bioRxiv, 2022: 2011-2022[2025-10-21]. .
[28]	WANG Y, GONG X, LI S, et al.. xTrimoABFold: de novo antibody structure prediction without MSA[J/OL]. arXiv, 2022: 2212.00735[2025-09-30]. .
[29]	ABANADES B, GEORGES G, BUJOTZEK A, et al.. ABlooper: fast accurate antibody CDR loop structure prediction with accuracy estimation[J]. Bioinformatics, 2022, 38(7): 1877-1880.
[30]	RUFFOLO J A, SULAM J, GRAY J J. Antibody structure prediction using interpretable deep learning[J/OL]. Patterns, 2022, 3(2): 100406[2025-12-21]. .
[31]	RUFFOLO J A, CHU L S, MAHAJAN S P, et al.. Fast, accurate antibody structure prediction from deep learning on massive set of natural antibodies[J/OL]. Nat. Commun., 2023, 14(1): 2389[2025-12-21]. .
[32]	LEE J H, YADOLLAHPOUR P, WATKINS A, et al.. EquiFold: protein structure prediction with a novel coarse-grained structure representation[J/OL]. bioRxiv, 2022[2025-10-21]. .
[33]	EVANS R, O'NEILL M, PRITZEL A, et al.. Protein complex prediction with AlphaFold-multimer[J/OL]. bioRxiv, 2021: 2015-2021[2025-10-21]. .
[34]	ABRAMSON J, ADLER J, DUNGER J, et al.. Accurate structure prediction of biomolecular interactions with AlphaFold 3[J]. Nature, 2024, 630(8016): 493-500.
[35]	JUMPER J, EVANS R, PRITZEL A, et al.. Highly accurate protein structure prediction with AlphaFold[J]. Nature, 2021, 596(7873): 583-589.
[36]	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.
[37]	JOUBBI S, MACCARI G, CIANO G, et al.. Improving antibody-antigen interaction prediction through flexibility with ESMFold[C]// Proceedings of the 18th International Joint Conference on Biomedical Engineering Systems and Technologies. SCITEPRESS-Science and Technology Publications, 2025: 603-610.
[38]	JOUBBI S, MICHELI A, MILAZZO P, et al.. Antibody design using deep learning: from sequence and structure design to affinity maturation[J/OL]. Brief. Bioinform., 2024, 25(4): bbae307[2025-09-30]. .
[39]	LI M, KANG L, XIONG Y, et al.. SESNet: sequence-structure feature-integrated deep learning method for data-efficient protein engineering[J/OL]. J. Cheminf., 2023, 15(1): 12[2025-12-21]. .
[40]	SHIN J E, RIESSELMAN A J, KOLLASCH A W, et al.. Protein design and variant prediction using autoregressive generative models[J/OL]. Nat. Commun., 2021, 12(1): 2403[2025-12-21]. .
[41]	HU R, FU L, CHEN Y, et al.. Protein engineering via Bayesian optimization-guided evolutionary algorithm and robotic experiments[J/OL]. Brief. Bioinform., 2023, 24(1): bbac570[2025-09-30]. .
[42]	KHAN A, COWEN-RIVERS A I, GROSNIT A, et al.. Toward real-world automated antibody design with combinatorial Bayesian optimization[J/OL]. Cell Rep. Methods, 2023, 3(1): 100374[2025-09-30]. .
[43]	FERRUZ N, SCHMIDT S, HÖCKER B. ProtGPT2 is a deep unsupervised language model for protein design[J/OL]. Nat. Commun., 2022, 13(1): 4348[2025-12-21]. .
[44]	CHUNGYOUN M, RUFFOLO J, GRAY J. FLAb: benchmarking deep learning methods for antibody fitness prediction[J/OL]. bioRxiv, 2024: 2021-2024[2025-10-21]. .
[45]	LIAO Y, MA H, WANG Z, et al.. Rapid restoration of potent neutralization activity against the latest Omicron variant JN.1 via AI rational design and antibody engineering[J/OL]. Proc. Natl. Acad. Sci. USA, 2025, 122(6): e2406659122[2025-09-30]. .
[46]	GUO P, LI M, PAN H, et al.. Multi-modality representation learning for antibody-antigen interactions prediction[C]// 2025 IEEE International Conference on Multimedia and Expo (ICME). Piscataway, New Jersey: IEEE, 2025: 1-6.
[47]	HOSSAIN D, SAGHAPOUR E, SONG K, et al.. Llama-affinity: a predictive antibody antigen binding model integrating antibody sequences with Llama3 backbone architecture[EB/OL]. arXiv, 2025: 2506.09052[2025-09-30]. .
[48]	AGARWAL A A, HARRANG J, NOBLE D, et al.. AlphaBind, a domain-specific model to predict and optimize antibody-antigen binding affinity[J/OL]. MAbs, 2025, 17(1): 2534626[2025-12-21]. .
[49]	ROLLINS Z A, WIDATALLA T, CHENG A C, et al.. AbMelt: learning antibody thermostability from molecular dynamics[J]. Biophys. J., 2024, 123(17): 2921-2933.
[50]	MAKOWSKI E K, KINNUNEN P C, HUANG J, et al.. Co-optimization of therapeutic antibody affinity and specificity using machine learning models that generalize to novel mutational space[J/OL]. Nat. Commun., 2022, 13(1): 3788[2025-09-30]. .
[51]	ZHOU X, XUE D, CHEN R, et al.. Antigen-specific antibody design via direct energy-based preference optimization[EB/OL]. arXiv, 2024: 2403.16576[2025-09-30]. .

名称	类别	模型	特点	参考文献
AbLang	抗体	蛋白质语言模型	通过改进的序列-功能关系捕捉能力执行抗体语言建模	[5]
AntiBERTy	抗体	蛋白质语言模型	基于BERT架构实现抗体序列的全面理解	[6]
IgLM	抗体	蛋白质语言模型	生成具有更高质量和多样性的抗体序列	[7]
nanoBERT	纳米抗体	蛋白质语言模型	针对骆驼科VHH结构域设计的专用纳米抗体语言模型	[8]
ProGen2-OAS	抗体	蛋白质语言模型	针对条件序列生成优化的抗体设计模型	[9]
IgBert	抗体	蛋白质语言模型	专用于免疫球蛋白序列分析与分类的BERT模型	[10]
IgT5	抗体	蛋白质语言模型	基于T5架构的抗体模型，支持翻译、生成和重构任务	[10]
GearBind	抗体	图神经网络	通过几何注意力机制预测抗体-抗原结合	[11]
RefineGNN	抗体	图神经网络	使用图神经网络优化抗体序列以提高结合能力	[12]
ProteinMPNN	通用	图神经网络	基于3D结构条件设计的蛋白质逆向折叠模型（最先进技术）	[13]
AbDiffuser	抗体	扩散模型	基于抗原结构条件生成抗体序列的扩散模型	[14]
RFdiffusion	通用	扩散模型	具备序列设计能力的蛋白质主链生成模型	[15]
Chroma	通用	扩散模型	支持结构约束的高分辨率全新蛋白质设计	[16]

名称	类别	模型	特点	参考文献
AbLang	抗体	蛋白质语言模型	通过改进的序列-功能关系捕捉能力执行抗体语言建模	[5]
AntiBERTy	抗体	蛋白质语言模型	基于BERT架构实现抗体序列的全面理解	[6]
IgLM	抗体	蛋白质语言模型	生成具有更高质量和多样性的抗体序列	[7]
nanoBERT	纳米抗体	蛋白质语言模型	针对骆驼科VHH结构域设计的专用纳米抗体语言模型	[8]
ProGen2-OAS	抗体	蛋白质语言模型	针对条件序列生成优化的抗体设计模型	[9]
IgBert	抗体	蛋白质语言模型	专用于免疫球蛋白序列分析与分类的BERT模型	[10]
IgT5	抗体	蛋白质语言模型	基于T5架构的抗体模型，支持翻译、生成和重构任务	[10]
GearBind	抗体	图神经网络	通过几何注意力机制预测抗体-抗原结合	[11]
RefineGNN	抗体	图神经网络	使用图神经网络优化抗体序列以提高结合能力	[12]
ProteinMPNN	通用	图神经网络	基于3D结构条件设计的蛋白质逆向折叠模型（最先进技术）	[13]
AbDiffuser	抗体	扩散模型	基于抗原结构条件生成抗体序列的扩散模型	[14]
RFdiffusion	通用	扩散模型	具备序列设计能力的蛋白质主链生成模型	[15]
Chroma	通用	扩散模型	支持结构约束的高分辨率全新蛋白质设计	[16]

名称	模型	优势	局限	应用	参考文献
ABodyBuilder2	AF-Multimer	减少OpenMM物理约束问题	CDR H3结构预测及物理合理结构生成仍具挑战	Fv及纳米抗体结构预测	[26]
tFold-Ab	AF-Multimer	结合蛋白质语言模型与AF-Multimer	PLM选择需进一步研究	Fv及纳米抗体结构预测	[27]
xTrimoABFold	AF-Multimer	采用AntiBERTy嵌入与快速模板搜索算法	未考虑复合物结构优化	Fv及纳米抗体结构预测	[28]
ABlooper	E(n)-EGNNs	无需外部工具/MSA，速度快	可能产生非物理预测	CDR环结构预测	[29]
DeepAb	LSTM+residual NN	支持点突变建议	依赖Rosetta（速度较慢）	Fv结构预测	[30]
IgFold	AntiBERTy+Graph transformer+IPA	AntiBERTy嵌入减少物理约束	依赖Rosetta生成主干结构	Fv及纳米抗体结构预测	[31]
EquiFold	SE(3)-equivariant NN	无需MSA/PLM，参数少速度快	可能产生非物理预测	迷你蛋白设计及Fv结构预测	[32]

名称	模型	优势	局限	应用	参考文献
ABodyBuilder2	AF-Multimer	减少OpenMM物理约束问题	CDR H3结构预测及物理合理结构生成仍具挑战	Fv及纳米抗体结构预测	[26]
tFold-Ab	AF-Multimer	结合蛋白质语言模型与AF-Multimer	PLM选择需进一步研究	Fv及纳米抗体结构预测	[27]
xTrimoABFold	AF-Multimer	采用AntiBERTy嵌入与快速模板搜索算法	未考虑复合物结构优化	Fv及纳米抗体结构预测	[28]
ABlooper	E(n)-EGNNs	无需外部工具/MSA，速度快	可能产生非物理预测	CDR环结构预测	[29]
DeepAb	LSTM+residual NN	支持点突变建议	依赖Rosetta（速度较慢）	Fv结构预测	[30]
IgFold	AntiBERTy+Graph transformer+IPA	AntiBERTy嵌入减少物理约束	依赖Rosetta生成主干结构	Fv及纳米抗体结构预测	[31]
EquiFold	SE(3)-equivariant NN	无需MSA/PLM，参数少速度快	可能产生非物理预测	迷你蛋白设计及Fv结构预测	[32]

名称	模型	优势	局限	应用
AlphaFold2(AF2)	Evoformer模块+端到端优化	框架区预测精度高；整体建模接近实验分辨率	CDR-H3预测受限于MSA稀疏性；依赖进化信息	单体蛋白预测、全蛋白组建模
AlphaFold3(AF3)	PairFormer+扩散模型，多链复合物预测	显著提升复合物、抗体-抗原界面与配体/核酸结合位点的预测精度；在界面残基取向和距离判断上优于AF2	计算成本较高；部分复杂体系预测仍不稳定	高精度复合物建模、抗体-抗原相互作用预测
ESM-2	大规模语言模型(15B 参数)，无需 MSA	序列→结构预测快速；支持大规模抗体库建模	对环区/复合物预测精度有限	快速抗体结构生成、高通量筛选
ESM-3	多模态生成模型(序列-结构-功能联合建模）	可设计新型功能蛋白；支持可控生成	模型参数巨大，资源需求高；完整版本未全面开源	抗体功能优化、序列-结构-功能一体化设计

抗体药物智能开发的深度学习策略

Deep Learning Strategies for Intelligent Development of Antibody Drugs

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 4

参考文献 51

相关文章 3

编辑推荐

Metrics

本文评价

[1]	李明, 曲楠楠, 李伟. 抗体药物连续生物工艺的技术突破与监管挑战[J]. 生物技术进展, 2025, 15(5): 845-853.
[2]	杨懿祺, 张志高, 游小龙, 张婧, 林冠峰, 吴英松. 抗体药物的发展与应用[J]. 生物技术进展, 2022, 12(3): 358-365.
[3]	辛志奇, 赵航, 汪海, 路铁刚. 基于深度学习的作物基因组学和遗传改良[J]. 生物技术进展, 2021, 11(4): 483-488.