目的 借助深度学习与基于结构的药物设计策略，开发兼具高效激动活性与酶切耐受性的胰高血糖素样肽-1受体（glucagon-like-peptide-1 receptor，GLP-1R）/胃抑制多肽受体（gastric inhibitory polypeptide receptor，GIPR）/胰高血糖素受体（glucagon receptor，GCGR）三受体激动多肽并评估其体外效力。方法 以临床候选多肽Peptide 20（MAR423）为模板，对N端前3残基进行虚拟饱和突变，采用Protein Message-passing Neural Network对三受体复合物稳定性评分，并结合神经网络模型预测二肽基肽酶-Ⅳ（dipeptidyl peptidase Ⅳ, DPP-Ⅳ）敏感性，筛选800条候选序列；对优选序列执行5 ns分子动力学模拟计算结合自由能，并在HEK-293报告基因体系中测定半数效应浓度（median effective concentration，EC₅₀）；耐酶性能通过DPP-Ⅳ处理后残余活性验证。结果 获得的Peptide A对GLP-1R、GIPR、GCGR的结合自由能分别达到（﹣151.49±15.87）、（﹣122.79±23.42）和（﹣139.25±19.81） kcal·mol^﹣¹，构象稳定；体外EC₅₀分别为0.069、0.089、2.730 nmol/L，与各内源性配体相当或更优。经DPP-Ⅳ处理24 h后活性保留率（104.30±3.23）%，显著优于对照肽（40.91%~89.00%）。结论 建立了快速、有效获得具有优异三受体平衡活性和耐DPP-Ⅳ能力新型多肽的流程，为代谢性疾病长效多靶点肽药物的理性设计提供了通用方案和实验依据。

Objective To develop a glucagon-like-peptide-1 receptor (GLP-1R)/gastric inhibitory polypeptide receptor (GIPR)/glucagon receptor (GCGR) triagonist peptide with balanced activity and strong resistance to dipeptidyl peptidase Ⅳ(DPP-Ⅳ) degradation using deep learning and structural modeling.Methods Based on clinical candidate Peptide 20 (MAR423), virtual mutagenesis was performed on the first 3 residues at the N-terminus， trireceptor complex stability was scored using Protein Message-passing Neural Network, and 800 sequences were screened based on the score and DDP-Ⅳ sensitivity predicted by neural network model. Top candidates were evaluated by 5 ns molecular dynamics simulations to calculate binding energy, and the median effective concentration (EC₅₀) was measured with HEK-293 reporter gene assay. DPP-Ⅳ resistance was assessed by measuring residual activity after enzyme treatment.Results Peptide A, modified with γGlu-C16, had stable conformation, with binding energy values to GLP-1R, GIPR, GCGR at (﹣151.49 ± 15.87), (﹣122.79 ± 23.42), (﹣139.25 ± 19.81) kcal·mol^﹣¹，respectively, and showed strong and balanced activity with EC₅₀ of 0.069, 0.089, 2.730 nmol/L, respectively. It retained (104.30±3.23)% activity after 24 h of DPP-Ⅳ treatment.Conclusion A rapid and effective strategy to discover triagonist peptide with improved stability and balanced potency is presented, supporting future development of multi-target peptide drugs for metabolic diseases.