Denenicokin

作者：冯景帆 薛超然 美工：何国红 罗真真 排版：马超 免疫原性（Immunogenicity）指外源性或内源性物质诱导机体产生适应性免疫应答的能力。在生物医药领域，抗体药物的免疫原性是指抗体药物及其相关成分在患者体内引发免疫应答的潜力，这种免疫反应可能导致一系列不良后果。首先，患者可能产生抗药物抗体（ADA），这是最常见的免疫反应，ADA的产生可能改变药物的药代动力学和药效学，从而影响疗效。其次，免疫反应还可能引发严重的临床不良反应，包括过敏反应、输液反应或自身免疫反应，危及患者安全。此外，ADA还可能中和抗体药物的活性，导致药物疗效降低甚至失效。因此，在抗体药物的设计、开发和监管过程中，免疫原性始终是一个不容忽视的重要因素。 ▲ 图1. 免疫原性对药物活性的影响 一、抗体免疫原性影响因素 抗药抗体（ADA）是抗体药物免疫原性评价的主要方式。ADA的形成可以分为与患者相关的因素和与药物相关的因素。 01 患者相关因素 患者个体差异是抗体免疫原性的重要决定因素，主要包括遗传背景、疾病状态和给药策略等。 ▶ 1) 遗传背景方面，主要组织相容性复合体（MHC）及人类白细胞抗原（HLA）等位基因的多样性显著影响抗药物抗体（ADA）的产生机制。研究表明，携带HLA-DRb-11、HLA-DQ-03、HLA-DQ-05等位基因的患者更易通过T细胞依赖性途径产生ADA，且不同人种因遗传异质性可能对同一药物的免疫原性响应存在差异。 ▶ 2) 疾病状态与免疫特征同样影响ADA生成。既往微生物或病毒感染史可能改变个体的免疫反应阈值，而年龄相关因素（如儿童与成人蛋白质代谢速率的差异）亦可导致免疫原性表现不同。此外，患者既往接触同类或相似单克隆抗体的治疗史可能通过免疫记忆效应增强后续治疗的ADA风险。 ▶ 3) 给药策略的优化对免疫原性调控至关重要。相较于单次给药，重复或间歇性给药方案显著增加ADA诱导概率，而联合免疫抑制剂（如甲氨蝶呤）可通过抑制适应性免疫应答降低ADA发生率。 02 药物相关因素 在药物特性层面，对免疫原性的影响主要体现在分子设计、外部原因造成的结构改变和糖基化等方面。 ▶ 1) 抗体分子设计是免疫原性的核心决定因素。尽管抗体人源化技术降低了非人源序列的免疫原性风险，但互补决定区（CDR）仍可能作为抗原表位触发抗独特型抗体（Anti-idiotypic Antibodies）的产生。 ▶ 2) 制剂工艺与储存条件通过改变蛋白质构象或引入杂质间接影响免疫原性。例如，容器封闭系统的相容性不足可能导致蛋白质聚集或脱酰胺化，而配方中的辅料杂质可能发挥佐剂效应激活先天免疫通路。 ▶ 3) 糖基化修饰模式的双向调控作用值得关注：保守糖基化可屏蔽蛋白质主干表位降低免疫原性，而异常糖型（如非哺乳动物表达系统引入的异源糖链）可能通过Toll样受体（TLR）激活固有免疫信号。此外，生产工艺差异（如表达系统选择、纯化残留物）、给药途径及剂量频率等因素均可能通过改变分子稳定性或暴露表位影响免疫原性结局。 二、抗体免疫原性产生机制 抗药抗体（ADA）的产生机制主要分为两大类：T细胞依赖性途径和非T细胞依赖性途径。在T细胞依赖性途径中，治疗性蛋白（如抗体或融合蛋白）的免疫原性主要源自抗原递呈细胞（例如未成熟树突状细胞和B细胞）对抗体的摄取，进而触发T细胞特异性免疫反应。非T细胞依赖性通路通过B细胞直接识别抗体药物活化并产生ADA。 ▶ 1) 在T细胞依赖性通路中，抗体药物作为外源蛋白被抗原呈递细胞（APC）摄取后，经蛋白酶体加工为线性表位，并通过MHC-II分子呈递至CD4⁺ T细胞表面。当T细胞受体（TCR）识别表位并接受共刺激信号（如CD28/B7）后，CD4⁺ T细胞活化并分泌细胞因子（如IL-4、IL-21），驱动B细胞分化为浆细胞并分泌抗药物抗体（ADA）。此外，活化的CD4⁺ T细胞可通过激活CD8⁺ T细胞等效应细胞，放大免疫应答强度。 ▶ 2) 非T细胞依赖性通路则由B细胞主导，其表面BCR直接识别抗体药物的构象表位，触发B细胞自主活化并产生ADA。此类应答无需T细胞辅助，但通常强度较弱且缺乏免疫记忆。 ▲ 图2. 免疫源性产生的分子机制 三、抗体免疫原性评价 对药品进行免疫原性风险分析，继而通过蛋白质工程来定向去除可能引起免疫原性的位点，成为降低免疫原性风险的一个主要方式。为系统性降低治疗性蛋白的免疫原性风险，现代生物药研发已有多种分析框架，覆盖从计算模拟预测到临床验证等多个方面。 01 计算模拟筛选 基于人工智能与生物信息学算法，计算机模拟可高效预测分子中潜在的免疫原性风险位点，比如T细胞表位预测和理化特性评估等。对于T细胞表位预测，可以通过MHC-I/II结合亲和力分析（如NetMHC、IEDB工具）识别T细胞表位，评估其与常见HLA等位基因（如HLA-DRB1*07:01）的相互作用风险。对于理化特性评估，可以预测易聚集区域（如TANGO算法）、化学修饰敏感位点（脱酰胺、氧化热点），以及非人源序列的同源性差异。此类分析可在分子设计早期识别高风险候选序列，指导CDR区人源化、表位掩蔽或糖基化位点插入等工程策略。 02 体外实验验证 根据分析方式的不同，分为非机制依赖性分析和机制依赖性分析两种。 对于非机制依赖性分析（No MOA Effects），可以通过MHC相关多肽质谱法MAPPS（MHC-Associated Peptide Proteomics）和T细胞增殖实验等方式进行检测。MAPPS是通过质谱鉴定APC加工后实际呈递的肽段，验证计算预测的T细胞表位。T细胞增殖实验是将药物裂解肽段与健康供体PBMC共培养，检测特异性T细胞活化（如CFSE稀释法、ELISpot检测IFN-γ分泌）。 对于机制依赖性分析（Potential MOA Effects），可以通过树突状细胞（DC）激活实验和抗原呈递细胞（APC）功能评估等方式进行检测。DC激活实验检测药物是否通过TLR/细胞因子信号激活DC（如CD86/CD83表达上调）。抗原呈递细胞（APC）功能评估是评估药物对MHC-II分子表达及共刺激信号的影响。 03 生物物理与先天免疫分析 生物物理与先天免疫分析可以通过生物物理特性表征、全血先天免疫实验和TLR报告细胞系实验等方式进行检测。 生物物理特性表征是通过SEC-HPLC（聚集体分析）、CE-SDS（电荷异质性）、糖基化质谱等技术，监控可能增强免疫原性的理化缺陷。全血先天免疫实验是检测药物在人体全血中诱导的细胞因子风暴风险（如IL-6、TNF-α释放）。TLR报告细胞系实验是利用表达TLR2/4/9的HEK-Blue细胞系，量化药物对TLR通路的激活强度。 04 体内/离体模型模拟 体内和离体模型模拟可以通过人工淋巴结模型（ALINE）、人源化小鼠模型和器官芯片（Organ-on-a-Chip）等方式进行检测。 ALINE是在体外3D培养系统中模拟T/B细胞互作，预测IgG型ADA的产生动力学。人源化小鼠模型是对小鼠移植人类免疫细胞（如HIS小鼠）或表达人类MHC分子（如HLA转基因小鼠），评估药物在类人免疫环境中的应答。Organ-on-a-Chip通过整合微流控与类器官技术，可以动态监测药物在局部组织（如肠道、皮肤）的免疫激活效应。 05 临床阶段监测与验证 临床阶段监测与验证可以通过ADA/nAb动态监测、免疫表型关联分析和表位再刺激验证等方式进行检测。 ADA/nAb动态监测是通过桥接ELISA、表面等离子共振（SPR）等技术，追踪患者血清中抗药物抗体（ADA）及中和抗体（nAb）的出现与滴度变化。免疫表型关联分析通过结合患者HLA分型、基线免疫细胞亚群（如Treg/Th17比例）及细胞因子谱（如IL-2、IL-10），解析个体差异对免疫原性的影响。表位再刺激验证通过从临床样本中分离T细胞，用药物衍生肽段再刺激，确认表位预测的临床相关性。 四、抗体免疫原性预测方法 下文将通过具体案例，阐释如何利用计算工具与实验平台的协同整合，实现从风险预测到分子优化的闭环设计，最终推动低免疫原性治疗抗体的临床转化。 01 ISPRI：免疫原性风险评估的云端智能平台 ISPRI（互动筛选与蛋白质重组界面）是由EpiVax公司开发的云端工具包，专注于生物治疗药物的免疫原性风险预测与分子优化。其核心算法EpiMatrix通过机器学习扫描蛋白质序列中的潜在T细胞表位，并评估其与人类白细胞抗原（HLA）的结合亲和力——该参数直接决定抗原呈递效率及免疫原性强度。平台创新性整合Tregitope分析模块，量化调节性T细胞表位诱导的免疫耐受效应，精准预测药物引发免疫激活或抑制的倾向性。针对高风险表位，OptiMatrix工具可智能生成序列优化策略，包括氨基酸替换、糖基化修饰或构象掩蔽，有效消除免疫原性热点。 ▲ 图3. ISPRI免疫源性预测结果 图3列出了一系列抗体，并标注了它们实际用ISPRI观察到的ADA反应百分比。纵轴（Y轴）：Tregitope-adjusted EpiMatrix Score表示经过Tregitope调整的EpiMatrix免疫原性评分。Tregitope是一些可以诱导免疫耐受的表位，调整后的评分考虑了这些表位的影响。纵轴左侧的数字显示的是Tregitope-adjusted EpiMatrix评分数值，右侧的括号内显示的是根据评分预测的ADA反应的百分比。将不同的抗体按照预测的EpiMatrix评分排列。图表中的抗体：图表中列出了一系列抗体，并标注了它们实际观察到的ADA反应百分比（例如，Alemtuzumab (45% obs ADA)）。图表左侧对免疫原性抗体和非免疫原性抗体进行了标注。Predicted ADA (±5%)：显示了根据EpiMatrix评分预测的ADA反应百分比，以及一个±5%的误差范围。Observed ADA：显示了实际临床观察到的ADA反应百分比。 02 IEDB：免疫表位数据库的构建与应用 免疫表位数据库(IEDB)是由美国国家过敏和传染病研究所（NIAID）支持的权威免疫学数据库，致力于系统化整合全球已验证的免疫表位实验数据。其通过人工审核与自动化文本挖掘技术，从海量科学文献中提取抗体表位、T细胞表位及MHC分子结合特性等关键信息，并进行标准化注释与结构化存储。 平台内嵌的免疫表位分析工具集（IEDB Analysis Resource）提供多项核心功能：基于人工神经网络（NetMHCpan）和矩阵模型（SMM）预测肽段与MHC-I/II分子的结合能力；利用BepiPred算法识别线性或构象型B细胞表位；结合表位保守性、HLA群体覆盖率和宿主交叉反应性数据，评估候选抗原的临床免疫原性风险等级。 ▲ 图4. IEDB免疫源性预测-B细胞空间表位预测 图4展示了IEDB对B细胞空间表位预测的结果，在B细胞表位预测分析中，阳性结果（高于设定阈值，以红色参考线为界）通过绿色标识，阴性结果则以橙色区分。3D结构视图通过Jmol工具展示，阳性预测的残基以黄色高亮标记，并支持鼠标交互（旋转、缩放及残基侧链显示）。结构分析数据以表格形式同步呈现，包括表位残基的链标识符（Chain ID）、残基编号（Residue ID）、接触数、倾向性评分及DiscoTope评分。点击表格中任一残基的CPK按钮，可在3D视图中切换至CPK模式并聚焦该位点的空间构象。 03 BioPhi：开源抗体工程平台的技术革新与应用价值 BioPhi是由国际科研团队开发的开源抗体设计平台，旨在通过人工智能与大数据技术优化治疗性抗体的开发流程。该平台整合两大核心模块——Sapiens（抗体人源化工具）与OASis（人源性评估系统），旨在解决传统抗体工程中的人源化效率低、免疫原性风险不可控等核心挑战。Sapiens基于深度神经网络架构，采用自然语言处理（NLP）中的语言模型技术，在Observed Antibody Space (OAS) 数据库上训练。OASis通过分析抗体序列与OAS数据库的局部相似性，评估人源性并预测免疫原性风险。 ▲ 图5. BioPh免疫源性预测 图5展示了BioPhi的免疫原性预测结果，OASis Identity（OASis相似度）: 表示抗体序列与Observed Antibody Space (OAS) 数据库中人类抗体序列的相似程度。OASis是BioPhi提供的一种人源性评分方法。数值越高，表示与人类抗体序列越相似，人源性越高。OASis Percentile（OASis百分位数）: 表示抗体序列在OAS数据库中的人源性排名。百分位数越低，表示人源性越低。图中的红色条形图直观的显示了百分位数的大小。Germline Content（种系含量）: 表示抗体序列与人类种系基因的相似程度。种系基因是未经体细胞突变的原始抗体基因。较高的种系含量通常意味着较低的免疫原性风险。Germline Gene（种系基因）: 显示了与分析的抗体序列最接近的种系基因的名称。 五、总结展望 随着生物技术的飞速发展，抗体药物、基因治疗和细胞治疗等新型疗法不断问世，免疫原性已成为影响其安全性与疗效的关键因素。如何有效预测、评估并控制免疫原性，正成为生物医药研发的重要挑战与新机遇。展望未来，三优生物正在构建一个集免疫原性预测、评估与改造于一体的智能化平台。该平台将通过构建全面的免疫原性数据库和智能化预测程序，实现对抗体批量免疫原性的精准预测并标注潜在的免疫原性肽段。对于高免疫原性的抗体，平台将提供线上改造方案，在不影响亲和力的前提下，对其免疫原性肽进行破除，从而实现免疫原性的线上改造。这将极大地提高抗体药物的研发效率和安全性，最终惠及广大患者。 Sanyou 10th Anniversary: Comprehensive Analysis of the Immunogenicity of Antibody Drugs Immunogenicity refers to the ability of exogenous or endogenous substances to induce an adaptive immune response in the body. In the field of biomedicine, the immunogenicity of antibody drugs refers to the potential of the antibody drug and its related components to trigger an immune response in patients. This immune response may lead to a series of adverse consequences. Firstly, patients may develop Anti-Drug Antibodies (ADA), the most common immune reaction. The production of ADA can alter the drug's pharmacokinetics and pharmacodynamics, thereby affecting efficacy. Secondly, the immune response may also trigger serious clinical adverse reactions, including allergic reactions, infusion reactions, or autoimmune reactions, jeopardizing patient safety. Additionally, ADA may neutralize the activity of the antibody drug, leading to reduced efficacy or even treatment failure. Therefore, immunogenicity is always a critical factor that cannot be ignored in the design, development, and regulation of antibody drugs. ▲ Figure 1. Impact of Immunogenicity on Drug Activity I. Factors Influencing Antibody Immunogenicity Anti-Drug Antibodies (ADA) are the primary method for evaluating the immunogenicity of antibody drugs. The formation of ADA can be divided into patient-related factors and drug-related factors. 01 Patient-Related Factors Individual patient differences are significant determinants of antibody immunogenicity, mainly including genetic background, disease status, and dosing strategy. ▶ 1) Genetic Background: The diversity of Major Histocompatibility Complex (MHC) and Human Leukocyte Antigen (HLA) alleles significantly influences the mechanisms of ADA production. Studies show that patients carrying alleles such as HLA-DRb-11, HLA-DQ-03, HLA-DQ-05 are more prone to producing ADA via T-cell-dependent pathways. Furthermore, due to genetic heterogeneity, different ethnic groups may exhibit varying immunogenic responses to the same drug. ▶ 2) Disease Status and Immune Characteristics: also affect ADA generation. A history of microbial or viral infections may alter an individual's immune response threshold, while age-related factors (e.g., differences in protein metabolism rates between children and adults) can also lead to different immunogenicity profiles. Additionally, a patient's prior exposure to the same or similar monoclonal antibodies may increase the risk of ADA in subsequent treatments through immune memory effects. ▶ 3) Dosing Strategy Optimization is crucial for immunogenicity control. Compared to single dosing, repeated or intermittent dosing regimens significantly increase the probability of ADA induction. Conversely, combination with immunosuppressants (e.g., methotrexate) can reduce ADA incidence by suppressing the adaptive immune response. 02 Drug-Related Factors Regarding drug characteristics, the impact on immunogenicity is mainly reflected in molecular design, structurally altered molecules caused by external factors, and glycosylation. ▶ 1) Antibody Molecular Design is a core determinant of immunogenicity. Although antibody humanization technology reduces the immunogenicity risk associated with non-human sequences, the Complementarity-Determining Regions (CDRs) can still act as epitopes and trigger Anti-idiotypic Antibodies. ▶ 2) Formulation Process and Storage Conditions indirectly affect immunogenicity by altering protein conformation or introducing impurities. For example, incompatibility with container closure systems may lead to protein aggregation or deamidation, while excipient impurities in the formulation might exert an adjuvant effect, activating innate immune pathways. ▶ 3) The Bidirectional Regulatory Role of Glycosylation Patterns warrants attention: conserved glycosylation can shield protein backbone epitopes, reducing immunogenicity, whereas aberrant glycoforms (e.g., heterologous glycans introduced by non-mammalian expression systems) may activate innate immune signaling through Toll-like Receptors (TLRs). Furthermore, differences in production processes (e.g., choice of expression system, purification residuals), route of administration, and dosing frequency may all influence immunogenicity outcomes by altering molecular stability or exposing epitopes. II. Mechanisms of Antibody Immunogenicity The mechanisms underlying Anti-Drug Antibody (ADA) production are primarily divided into two categories: T-cell-dependent pathways and non-T-cell-dependent pathways. In T-cell-dependent pathways, the immunogenicity of therapeutic proteins (such as antibodies or fusion proteins) mainly stems from the uptake of the antibody by antigen-presenting cells (e.g., immature dendritic cells and B cells), subsequently triggering T-cell-specific immune responses. Non-T-cell-dependent pathways involve the direct recognition of the antibody drug by B cells, leading to their activation and ADA production. ▶ 1) In the T-cell-dependent pathway, the antibody drug, as a foreign protein, is taken up by Antigen Presenting Cells (APCs), processed into linear epitopes by proteasomes, and presented on the surface of CD4⁺ T cells via MHC-II molecules. When the T-cell receptor (TCR) recognizes the epitope and receives co-stimulatory signals (e.g., CD28/B7), CD4⁺ T cells are activated and secrete cytokines (e.g., IL-4, IL-21), driving B cells to differentiate into plasma cells and secrete ADAs. Additionally, activated CD4⁺ T cells can amplify the immune response intensity by activating effector cells like CD8⁺ T cells. ▶ 2) The non-T-cell-dependent pathway is primarily mediated by B cells, where their surface BCR directly recognizes conformational epitopes of the antibody drug, triggering autonomous B cell activation and ADA production. This type of response does not require T-cell help but is generally weaker and lacks immune memory. ▲ Figure 2. Molecular Mechanisms of Immunogenicity Generation III. Evaluation of Antibody Immunogenicity Conducting immunogenicity risk analysis for a drug product, followed by using protein engineering to specifically remove potential immunogenic sites, has become a major strategy for mitigating immunogenicity risk. To systematically reduce the immunogenicity risk of therapeutic proteins, modern biopharmaceutical R&D employs various analytical frameworks covering computational prediction to clinical validation. 01 Computational Screening Leveraging artificial intelligence and bioinformatics algorithms, computer simulations can efficiently predict potential immunogenic risk sites within a molecule, such as T-cell epitope prediction and physiochemical property assessment. T-cell Epitope Prediction: Tools like NetMHC and IEDB analyze MHC-I/II binding affinity to identify T-cell epitopes and assess their interaction risk with common HLA alleles (e.g., HLA-DRB1*07:01). Physiochemical Property Assessment: Predicts aggregation-prone regions (using algorithms like TANGO), chemically modification-sensitive sites (deamidation, oxidation hotspots), and homology differences in non-human sequences. Such analyses can identify high-risk candidate sequences early in molecular design, guiding engineering strategies like CDR humanization, epitope masking, or glycosylation site insertion. 02 In Vitro Assay Validation Depending on the analysis method, this is divided into non-mechanism of action (Non-MOA) dependent analysis and mechanism of action (MOA) dependent analysis.  Non-MOA Dependent Analysis include MHC-Associated Peptide Proteomics (MAPPS): Uses mass spectrometry to identify peptides actually presented by APCs after processing, validating computationally predicted T-cell epitopes. Non-MOA also cover T-cell Proliferation Assay: Co-cultures drug cleavage peptides with healthy donor PBMCs to detect specific T-cell activation (e.g., using CFSE dilution assay, ELISpot for IFN-γ secretion). Potential MOA Dependent Analysis have Dendritic Cell (DC) Activation Assay: Detects whether the drug activates DCs via TLR/cytokine signaling (e.g., upregulation of CD86/CD83 expression). Also include Antigen Presenting Cell (APC) Function Assessment: Evaluates the drug's impact on MHC-II molecule expression and co-stimulatory signals. 03 Biophysical and Innate Immunity Analysis Biophysical Characterization: Uses techniques like SEC-HPLC (aggregate analysis), CE-SDS (charge heterogeneity), glycosylation mass spectrometry to monitor physicochemical defects that may enhance immunogenicity.  Whole Blood Innate Immunity Assay: Detects the risk of cytokine storm induced by the drug in human whole blood (e.g., IL-6, TNF-α release).  TLR Reporter Cell Line Assay: Utilizes HEK-Blue cell lines expressing TLR2/4/9 to quantify the drug's activation strength of TLR pathways. 04 In Vivo/Ex Vivo Model Simulation Artificial Lymph Node Model (ALINE): Simulates T/B cell interactions in an in vitro 3D culture system to predict the production kinetics of IgG-type ADA. Humanized Mouse Models: Involves mice engrafted with human immune cells (e.g., HIS mice) or expressing human MHC molecules (e.g., HLA transgenic mice) to assess drug response in a human-like immune environment. Organ-on-a-Chip: Integrates microfluidics and organoid technology to dynamically monitor the immune activation effects of drugs in local tissues (e.g., gut, skin). 05 Clinical Stage Monitoring and Validation ADA/nAb Dynamic Monitoring: Uses techniques like bridging ELISA, Surface Plasmon Resonance (SPR) to track the appearance and titer changes of ADAs and neutralizing antibodies (nAbs) in patient serum. Immune Phenotype Correlation Analysis: Combines patient HLA typing, baseline immune cell subsets (e.g., Treg/Th17 ratio), and cytokine profiles (e.g., IL-2, IL-10) to analyze the impact of individual differences on immunogenicity. Epitope Re-stimulation Validation: Isolates T cells from clinical samples and re-stimulates them with drug-derived peptides to confirm the clinical relevance of epitope predictions. IV. Methods for Predicting Antibody Immunogenicity The following section illustrates, through specific cases, how the synergistic integration of computational tools and experimental platforms enables closed-loop design from risk prediction to molecular optimization, ultimately facilitating the clinical translation of therapeutic antibodies with low immunogenicity. 01 ISPRI: Cloud-based Intelligent Platform for Immunogenicity Risk Assessment ISPRI (i, meaning interactive, is the naming convention of EpiVax) is a cloud-based toolkit developed by EpiVax Inc., focusing on immunogenicity risk prediction and molecular optimization for biotherapeutics. Its core algorithm, EpiMatrix, uses machine learning to scan protein sequences for potential T-cell epitopes and assesses their binding affinity to Human Leukocyte Antigens (HLAs) – a parameter directly determining antigen presentation efficiency and immunogenicity strength. The platform innovatively integrates the Tregitope analysis module, quantifying the immune tolerance effect induced by regulatory T cell epitopes, to accurately predict the drug's tendency to trigger immune activation or suppression. For high-risk epitopes, the OptiMatrix tool intelligently generates sequence optimization strategies, including amino acid substitution, glycosylation modification, or conformational masking, effectively eliminating immunogenicity hotspots. ▲ Figure 3. ISPRI Immunogenicity Prediction Results Figure 3 lists a series of antibodies and annotates their actual observed ADA response percentages using ISPRI. Y-axis: Tregitope-adjusted EpiMatrix Score. The numbers on the left side of the Y-axis show the Tregitope-adjusted EpiMatrix score values, while the parentheses on the right show the predicted percentage of ADA response based on the score. Different antibodies are arranged according to their predicted EpiMatrix score. Antibodies in the chart: The chart lists a series of antibodies and annotates their actual observed ADA response percentages (e.g., Alemtuzumab (45% obs ADA)). The left side of the chart annotates immunogenic and non-immunogenic antibodies. Predicted ADA (±5%): Shows the predicted ADA response percentage based on the EpiMatrix score, with a ±5% error range. Observed ADA: Shows the actual clinically observed ADA response percentage. 02 IEDB: Construction and Application of the Immune Epitope Database The Immune Epitope Database (IEDB) is an authoritative immunology database supported by the National Institute of Allergy and Infectious Diseases (NIAID), dedicated to systematically integrating globally validated experimental data on immune epitopes. Through manual curation and automated text mining, it extracts key information such as antibody epitopes, T-cell epitopes, and MHC molecule binding characteristics from vast scientific literature, followed by standardized annotation and structured storage. The embedded Immune Epitope Database Analysis Resource (IEDB Analysis Resource) provides several core functions: predicting peptide binding capacity to MHC-I/II molecules based on artificial neural networks (NetMHCpan) and matrix models (SMM); identifying linear or conformational B-cell epitopes using the BepiPred algorithm; assessing the clinical immunogenicity risk level of candidate antigens by combining epitope conservation, HLA population coverage, and host cross-reactivity data. ▲ Figure 4. IEDB Immunogenicity Prediction - B-cell Conformational Epitope Prediction Figure 4 shows the results of IEDB's prediction of B-cell conformational epitopes. In the B-cell epitope prediction analysis, positive results (above the set threshold, marked by the red reference line) are indicated in green, while negative results are distinguished in orange. The 3D structure view, displayed using the Jmol tool, highlights positive predicted residues in yellow and supports mouse interaction (rotation, zoom, and residue side chain display). Structural analysis data is presented simultaneously in table form, including the chain identifier (Chain ID), residue number (Residue ID), number of contacts, propensity score, and DiscoTope score for epitope residues. Clicking the CPK button for any residue in the table switches the 3D view to CPK mode and focuses on the spatial conformation of that site. 03 BioPhi: Technological Innovation and Application Value of an Open-Source Antibody Engineering Platform BioPhi is an open-source antibody design platform developed by an international research team, aiming to optimize the development process of therapeutic antibodies through artificial intelligence and big data technologies. The platform integrates two core modules – Sapiens (an antibody humanization tool) and OASis (a humanness assessment system) – designed to address core challenges in traditional antibody engineering such as low humanization efficiency and uncontrolled immunogenicity risk. Sapiens is based on a deep neural network architecture, employing language model technology from Natural Language Processing (NLP), trained on the Observed Antibody Space (OAS) database. OASis evaluates humanness and predicts immunogenicity risk by analyzing the local similarity of antibody sequences to the OAS database. ▲ Figure 5. BioPhi Immunogenicity Prediction Figure 5 shows the immunogenicity prediction results from BioPhi. OASis Identity: Represents the degree of similarity between the antibody sequence and human antibody sequences in the Observed Antibody Space (OAS) database. OASis is a humanness scoring method provided by BioPhi. A higher value indicates greater similarity to human antibody sequences, implying higher humanness. OASis Percentile: Represents the humanness ranking of the antibody sequence within the OAS database. A lower percentile indicates lower humanness. The red bar chart in the figure visually shows the size of the percentile. Germline Content: Represents the degree of similarity between the antibody sequence and human germline genes. Germline genes are the original, unmutated antibody genes. Higher germline content typically implies lower immunogenicity risk. Germline Gene: Shows the name of the germline gene most similar to the analyzed antibody sequence. V. Summary and Outlook  With the rapid development of biotechnology and the continuous emergence of novel therapies like antibody drugs, gene therapy, and cell therapy, immunogenicity has become a key factor affecting their safety and efficacy. How to effectively predict, evaluate, and control immunogenicity is becoming a major challenge and a new opportunity in biopharmaceutical R&D. Looking ahead, Sanyou Bio is building an intelligent platform integrating immunogenicity prediction, assessment, and modification. This platform will achieve accurate batch prediction of antibody immunogenicity and annotation of potential immunogenic peptides by constructing a comprehensive immunogenicity database and intelligent prediction programs. For antibodies with high immunogenicity, the platform will provide online modification solutions to disrupt their immunogenic peptides without affecting affinity, thus enabling online immunogenicity engineering. This will significantly improve the R&D efficiency and safety of antibody drugs, ultimately benefiting a wide range of patients. ▶ References: [1] Jawa V, et al. T-Cell Dependent Immunogenicity of Protein Therapeutics Pre-clinical Assessment and Mitigation-Updated Consensus and Review 2020. Front Immunol. 2020 Jun 30;11:1301. [2] Mattei AE, et al. In silico methods for immunogenicity risk assessment and human homology screening for therapeutic antibodies. MAbs. 2024 Jan-Dec;16(1):2333729. [3] Prihoda D, et al. BioPhi: A platform for antibody design, humanization, and humanness evaluation based on natural antibody repertoires and deep learning. MAbs. 2022 Jan-Dec;14(1):2020203. [4] Vita R, et al. The Immune Epitope Database (IEDB): 2018 update. Nucleic Acids Res. 2019 Jan 8;47(D1):D339-D343. 关于三优生物 三优生物是一家以“让天下没有难做的创新生物药”为使命，以超万亿分子库和智能科技驱动的生物医药高科技企业。 公司致力于打造全球顶尖的原创新药创新工场。公司以智能超万亿分子库（AI-STAL）为核心；以干湿结合、国际领先的创新生物药智能化及一体化研发平台为依托；以多样化的业务模式推动全球创新药物的研发及产业化。 公司总部位于中国上海，在亚洲、北美洲、欧洲等多地建立了业务中心，形成了全球化的业务网络，现有投产及布局的研发及GMP场地20000多平方米。 公司已与全球2000多家药企、生技公司等建立了良好的合作关系，已赋能1200多个新药研发项目；已完成50多个合作研发项目，其中10多个协同研发项目已推至IND及临床研发阶段。 公司已申请130多项发明专利，其中30多项发明专利已获得授权，并获得了国家级高新技术企业、上海市专精特新、ISO9001、ISO27001等10余项资质及体系认证。 1 推荐阅读 三优十周年｜肿瘤免疫2025及双诺奖解析 三优十周年｜TCE药物-重塑免疫治疗版图的新力量 三优十周年｜神经系统疾病治疗进展 三优十周年｜AI-STAL 智能超万亿分子库发展历程 三优十周年｜打造全球顶尖的原创新药创新工场 三优十周年｜抗体治疗50年风雨江湖 三优十周年｜针对宠物疾病的单克隆抗体药物 三优十周年｜CAR-T细胞治疗现状挑战与未来 三优十周年｜服务篇-体外药效筛选解决方案 三优十周年｜AI-STAL篇-Anticalin靶向蛋白库 三优十周年｜靶点篇-自身免疫疾病靶点介绍 创新前沿｜三优生物类器官产学研联盟启航 中和抗体“硬核出击”，斩断CHIKV病毒感染链 三优十周年｜服务篇-靶点至PCC药物研发 三优十周年｜靶点篇-六大恶性肿瘤靶点介绍创新前沿｜2025ADC热门靶点与技术创新启示