人工智能驱动的SARS-CoV-2膜蛋白多肽抑制剂从头设计与动态机制研究

AI-driven de novo design and dynamic mechanistic study of peptide inhibitors targeting the SARS-CoV-2 membrane protein

摘要

目的 运用人工智能辅助的计算生物学方法, 以SARS-CoV-2病毒的M蛋白作为靶标, 设计并筛选出具有高效靶向M蛋白潜力的多肽抑制剂。方法 采用基于深度学习的蛋白质生成模型RFDiffusion构建多肽的主链骨架, 并利用序列设计工具ProteinMPNN优化氨基酸序列, 通过AlphaFold3软件包进行三维结构建模及评估; 结合MM/PBSA方法计算结合自由能(ΔG_bind), 筛选出ΔG_bind最优的16条多肽; 进一步通过聚类分析选取了3条代表性多肽, 进行相互作用位点分析、拉氏图构象验证及1 000 ns分子动力学(MD)模拟, 系统评估其结合稳定性与动态相互作用特征。结果 成功获得了3条具有高亲和力潜质的多肽分子; 拉氏图分析表明其主链二面角均位于允许构象区域, 结构相对合理; MD模拟揭示了M蛋白与多肽之间以氢键和盐桥为主的关键相互作用。结论 建立了一套从“生成设计序列优化结构验证动态评估”的智能化多肽设计流程, 通过纯计算模拟方法获得了靶向SARS-CoV-2 M蛋白的高结合潜力多肽, 并初步阐明了其可能的结合机制, 为后续实验验证与抗病毒药物研发提供了理论依据与先导结构。

Abstract

Objective To design and screen peptide inhibitors that effectively target the M protein using the artificial intelligence (AI)-assisted computational biology and taking the M protein of the SARS-CoV-2 virus as the target. Methods We employed RFDiffusion, a deep learning-based protein structure generation model, to design peptide backbone scaffolds. Amino acid sequences were then optimized using ProteinMPNN method, and the three-dimensional structure modeling and evaluation were carried out by AlphaFold3. Binding free energies (ΔG _bind) were calculated using the MM/PBSA method, and used for the selection of 16 top-affinity peptides. Three representative peptides were further chosen through the cluster analysis for the in-depth characterization, including the analysis of interaction sites, validation of Ramachandran plot, and 1 000 ns molecular dynamics (MD) simulations, in order to assess binding stability and dynamic interaction profiles. Results Three high-affinity peptide candidates were successfully obtained. Ramachandran plot analysis indicated that the dihedral angles of the main chain were all within the allowed regions, and the structure was relatively reasonable. MD simulations revealed the key interaction between the M protein and the designed peptides, mainly involving the hydrogen bonds and salt bridges. Conclusion We have established an integrated AI-driven pipeline for peptide design, covering stages from generative scaffolding, sequence optimization, and structural validation to dynamic binding assessment. Through the prediction of computational simulations, high binding potential peptides targeting the M protein of SARS-CoV-2 are predicted, and the possible binding mechanism is preliminarily elucidated. This will provide theoretical basis and lead structure for subsequent experimental verification and the development of antiviral drugs.

关键词

SARS-CoV-2膜蛋白 / 多肽抑制剂 / 人工智能辅助药物设计 / 分子动力学模拟 / 结合自由能 / 蛋白质结构预测 / 分子对接 / 抗病毒药物设计

Key words

SARS-CoV-2 membrane protein / peptide inhibitors / artificial intelligence-aided drug design / molecular dynamics simulation / binding free energy / protein structure prediction / molecular docking / antiviral drug design

参考文献

[1] ZHANG Z K, NOMURA N, MURAMOTO Y, et al. Structure of SARS-CoV-2 membrane protein essential for virus assembly[J]. Nat Commun, 2022, 13(1): 4399. DOI: 10.1038/s41467-022-32019-3.

[2] DOLAN K A, DUTTA M, KERN D M, et al. Structure of SARS-CoV-2 M protein in lipid nanodiscs[J]. eLife, 2022, 11: e81702. DOI: 10.7554/eLife.81702.

[3] CHEN Y, MASON G H, SCOURFIELD D O, et al. Structural definition of HLA class II-presented SARS-CoV-2 epitopes reveals a mechanism to escape pre-existing CD4⁺ T cell immunity [J]. Cell Rep, 2023, 42(8): 112827. DOI: 10.1016/j.celrep.2023.112827.

[4] EL-MARADNY Y A, BADAWY M A, MOHAMED K I, et al. Unraveling the role of the nucleocapsid protein in SARS-CoV-2 pathogenesis: from viral life cycle to vaccine development[J]. Int J Biol Macromol, 2024, 279(Pt 2): 135201. DOI: 10.1016/j.ijbiomac.2024.135201.

[5] MOTHAE S A, CHILIZA T E, MVUBU N E. SARS-CoV-2 host-pathogen interactome: insights into more players during pathogenesis[J]. Virology, 2025, 610: 110607. DOI: 10.1016/j.virol.2025.110607.

[6] 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. DOI: 10.1038/s41586-023-06415-8.

[7] BENNETT N R, WATSON J L, RAGOTTE R J, et al. Atomically accurate de novo design of antibodies with RFdiffusion[J]. Nature, 2026, 649(8095): 183-193. DOI: 10.1038/s41586-025-09721-5.

[8] CASE D A, AKTULGA H M, BELFON K, et al. AmberTools[J]. J Chem Inf Model, 2023, 63(20): 6183-6191. DOI: 10.1021/acs.jcim.3c01153.

[9] JUMPER J, EVANS R, PRITZEL A, et al. Highly accurate protein structure prediction with AlphaFold[J]. Nature, 2021, 596(7873): 583-589. DOI: 10.1038/s41586-021-03819-2.

[10] ABRAMSON J, ADLER J, DUNGER J, et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3[J]. Nature, 2024, 630(8016): 493-500. DOI: 10.1038/s41586-024-07487-w.

[11] LEE J, CHENG X, SWAILS J M, et al. CHARMM-GUI input generator for NAMD, GROMACS, AMBER, OpenMM, and CHARMM/OpenMM simulations using the CHARMM36 additive force field[J]. J Chem Theory Comput, 2016, 12(1): 405-413. DOI: 10.1021/acs.jctc.5b00935.

[12] DUAN H C, HU K X, ZHENG D, et al. Recognition and release of uridine and hCNT3: from multivariate interactions to molecular design[J]. Int J Biol Macromol, 2022, 223(Pt A): 1562-1577. DOI: 10.1016/j.ijbiomac.2022.11.145.

[13] CASE D A, CERUTTI D S, CRUZEIRO V W D, et al. Recent developments in amber biomolecular simulations[J]. J Chem Inf Model, 2025, 65(15): 7835-7843. DOI: 10.1021/acs.jcim.5c01063.

[14] KARIM M R, BEYAN O, ZAPPA A, et al. Deep learning-based clustering approaches for bioinformatics[J]. Brief Bioinform, 2021, 22(1): 393-415. DOI: 10.1093/bib/bbz170.

[15] DAVIS I W, LEAVER-FAY A, CHEN V B, et al. MolProbity: all-atom contacts and structure validation for proteins and nucleic acids[J]. Nucleic Acids Res, 2007, 35(Web Server issue): W375-W383. DOI: 10.1093/nar/gkm216.

[16] SMARDZ P, ANILA M M, ROGOWSKI P, et al. A practical guide to all-atom and coarse-grained molecular dynamics simulations using amber and gromacs: a case study of disulfide-bond impact on the intrinsically disordered amyloid beta[J]. Int J Mol Sci, 2024, 25(12): 6698. DOI: 10.3390/ijms25126698.

[17] ROE D R, CHEATHAM T E 3rd. PTRAJ and CPPTRAJ: software for processing and analysis of molecular dynamics trajectory data[J]. J Chem Theory Comput, 2013, 9(7): 3084-3095. DOI: 10.1021/ct400341p.

[18] BOUYSSET C, FIORUCCI S. ProLIF: a library to encode molecular interactions as fingerprints[J]. J Cheminform, 2021, 13(1): 72. DOI: 10.1186/s13321-021-00548-6.

[19] BAISHYA T, DUTTA KK, FRONTERA A, et al. On the importance of H-bonding interactions in the enclathration of boric acids in Na(I) polymers: experimental and theoretical studies[J]. Crystals, 2023, 13(6): 895. DOI: 10.3390/cryst13060895.

[20] DAVE D R, KASSEM S, COSTE M, et al. Adaptive peptide dispersions enable drying-induced biomolecule encapsulation[J]. Nat Mater, 2025, 24(9): 1465-1475. DOI: 10.1038/s41563-025-02300-z.

[21] ZHAO J S, ZHAO Y Z, ZHANG S L, et al. Insights into the chirality-dependent recognition of Danshensu Bingpian Zhi stereoisomers with PPAR_γ [J]. Phys Chem Chem Phys, 2024, 26(44): 28143-28154. DOI: 10.1039/d4cp03926b.

[22] HOLLINGSWORTH S A, KARPLUS P A. A fresh look at the Ramachandran plot and the occurrence of standard structures in proteins[J]. Biomol Concepts, 2010, 1(3/4): 271-283. DOI: 10.1515/BMC.2010.022.

[23] FABIAN P, STAPOR K, BANACH M, et al. Different synergy in amyloids and biologically active forms of proteins[J]. Int J Mol Sci, 2019, 20(18): 4436. DOI: 10.3390/ijms20184436.

[24] AISWARYA P, JAYAVARTHANAN T, PERIANDY S, et al. Molecular structural analysis, conformers and spectral (FT-IR, FT-Raman, NMR and UV-Visible), importance of solvent role in molecular, ADME and molecular docking investigation on alpha-cyano-4-hydroxycinnamic acid[J]. Chem Phys Impact, 2023, 7: 100353. DOI: 10.1016/j.chphi.2023.100353.

[25] GOSWAMI V, PATEL D, ROHIT S, et al. Homology modeling, binding site identification, molecular docking and molecular dynamics simulation study of emerging and promising drug target of Wnt signaling-Human Porcupine enzyme[J]. Results Chem, 2024, 7: 101482. DOI: 10.1016/j.rechem.2024.101482.

[26] GASHTI A B, AGBAYANI G, HRAPOVIC S, et al. Production, purification and immunogenicity of Gag virus-like particles carrying SARS-CoV-2 components[J]. Vaccine, 2024, 42(1): 40-52. DOI: 10.1016/j.vaccine.2023.11.048.

[27] CHAN J F, YUAN S F, CHU H, et al. COVID-19 drug discovery and treatment options[J]. Nat Rev Microbiol, 2024, 22(7): 391-407. DOI: 10.1038/s41579-024-01036-y.

[28] VAN DAMME E, ABEYWICKREMA P, YIN Y T, et al. A small-molecule SARS-CoV-2 inhibitor targeting the membrane protein[J]. Nature, 2025, 640(8058): 506-513. DOI: 10.1038/s41586-025-08651-6.

[29] LIU Y F, WANG S, DONG J X, et al. De novo protein design with a denoising diffusion network independent of pretrained structure prediction models[J]. Nat Methods, 2024, 21(11): 2107-2116. DOI: 10.1038/s41592-024-02437-w.

[30] HU R E, YU C H, NG I S. GRACE: generative redesign in artificial computational enzymology[J]. ACS Synth Biol, 2024, 13(12): 4154-4164. DOI: 10.1021/acssynbio.4c00624.

[31] AHERN W, YIM J, TISCHER D, et al. Atom-level enzyme active site scaffolding using RFdiffusion2[J]. Nat Methods, 2026, 23(1): 96-105. DOI: 10.1038/s41592-025-02975-x.

[32] LAUKO A, PELLOCK S J, SUMIDA K H, et al. Computational design of serine hydrolases[J]. Science, 2025, 388(6744): eadu2454. DOI: 10.1126/science.adu2454.

[33] 段晓琼, 谢鹤, 陈利民. SARS-CoV-2与宿主固有免疫系统的相互作用[J]. 四川大学学报(医学版), 2022, 53(1): 1-6. DOI: 10.12182/20220160101.

[34] 马明仁, 马凌, 刘燕, 等. SARS-CoV-2靶向CypA/CD147受体途径诱导心肌细胞凋亡[J]. 西安交通大学学报(医学版), 2024, 45(5): 734-740. DOI: 10.7652/jdyxb202405006.

[35] 尹慧敏, 吕海, 迟莹, 等. 人源性抗SARS-CoV-2单链抗体文库的构建及广谱中和抗体的筛选与鉴定[J]. 细胞与分子免疫学杂志, 2025, 41(2): 154-160. DOI: 10.13423/j.cnki.cjcmi.009906.

基金资助

国家自然科学基金杰出青年科学基金(T2425029)

重大疾病新药靶发现及新药创制全国重点实验室开放课题(SKLD2025M04)