白塞氏综合征（Behçet’s Syndrome, BS）是一种具有不同临床特征的可变性血管炎，以口腔阿弗他溃疡、生殖器溃疡和葡萄膜炎为主要临床三联征，累及动静脉、胃肠道和神经系统时会造成患者机体的严重损伤与死亡。目前已有的BS疾病诊断标准是基于临床症状和皮肤黏膜表现的，仍缺乏有效的实验室诊断学指标。该病在活动复发和缓解期的动态临床表现、疾病表现谱的复杂交叉为筛选识别BS生物标志物带来巨大挑战。因此，亟需寻找与疾病鉴别诊断、表型划分和病程（活动度、严重度）相关，应用于早期诊断、治疗反应监测和病情评估等方面的生物标志物。本研究从血浆自身抗体、脑脊液-血液配对蛋白组学、实验室常用指标和基因组学出发，筛选和鉴定BS特异性生物标志物，构建BS诊断及临床分型的标志物模型，并探索特异性分子与BS发病机制的关联。本论文主要研究内容分为以下六个章节：

第一章 白塞氏综合征自身抗体的研究

白塞氏综合征（Behçet’s Syndrome, BS）缺乏反映临床表型异质性和疾病病程的自身抗体标志物，这阻碍了其标准化诊断和精准治疗。因此，我们利用高通量蛋白芯片平台来识别BS的广谱性及表型特异性自身抗体。本研究共收集了来自495名BS患者、227名疾病对照组和118名健康对照组的840份血浆样本，利用大芯片初筛-小芯片验证-ELISA确认的方案锚定BS自身抗体标志物。采用机器学习的方法构建并优化用于BS诊断的自身抗体组合。并通过生物信息学差异分析、通路富集和聚类算法，阐明了表型特异性以及与疾病严重程度相关的自身抗体在不同器官/组织损伤中的免疫病理作用与重要性。此外，我们对BS患者的病理切片进行了免疫组织化学染色，以检测自身抗体对应靶抗原的定位。最后采用孟德尔随机化的方法验证了靶抗原对应基因多态性的表达与BS临床表现之间的因果关联。首先，通过大芯片初步筛选，BS和健康对照组间的差异表达靶抗原在血小板脱颗粒、分子伴侣介导的自噬和中性粒细胞激活通路中显著富集。第二，我们利用小芯片验证队列，构建了包含19个候选自身抗体的机器学习模型来鉴别诊断BS和对照组，其受试者工作曲线下面积（Area Under the Curve, AUC）高达0.82。其次，我们鉴定出了抗HBS1L抗体是BS诊断的“广谱性”标志物，而抗PPP1R13L抗体在疾病严重程度评估中表现出一定潜力。此外，抗p017抗体是胃肠道受累BS的表型特异性自身抗体。最后，经过免疫组化验证，我们发现CCDC140主要在BS结肠组织的细胞质基质和内部粘液中表达，是导致胃肠道病变的自身抗体主要靶点，也能够用于区分胃肠道受累表型的BS患者与炎症性肠病对照。本研究中，我们利用自身抗体组学发现了BS存在从自身炎症性发病向自身免疫性疾病过渡的中间状态，为广谱性和表型性BS的临床诊断提供了潜在的自身抗体候选标志物。

第二章 神经白塞病脑脊液-血浆配对蛋白组学研究

神经白塞病（Neuro-Behçet’s disease, NBD）是较为严重的BS神经受累表型，临床上较为罕见，主要分为实质型和非实质型NBD。病理层面的周围血管炎伴有中枢神经系统的免疫细胞浸润，以及脑脊液IL-6水平升高是NBD患者的特征性神经炎症表现。目前尚无研究从高通量蛋白组学层面揭示NBD病程/发病机制相关的特征性蛋白类标志物。本研究利用数据非依赖型质谱对BS患者的脑脊液-血液配对样本进行了全面的蛋白组学筛选，初步分析了NBD和对照组患者在脑脊液与血浆层面差异表达的蛋白及其参与的生物学致病通路。在脑脊液维度：我们共鉴定到了117种差异蛋白，其主要参与补体激活和体液免疫应答通路。通过与人类蛋白图谱数据库比对，我们发现这些蛋白主要表达于大脑皮质、小脑和尾状核，并调控活性氧损伤应答通路。在血浆层面，共有79个差异蛋白在NBD组显著高表达，并作用于急性炎症、凝血、轴突发育及体液免疫通路；其中，主要在中枢神经系统表达的蛋白则通过调节内质网应激反应发挥生物学功能。在临床诊断价值层面，我们集中于14个同时在NBD患者脑脊液和血浆中差异表达的蛋白，以期发现“无创性”血浆学替代性标志物。我们初步发现脑脊液和血浆中VSIG4蛋白联合检测在区分NBD与对照组患者时的AUC为0.824。在NBD时序性样本的差异表达分析中，我们在脑脊液与血浆层面分别筛选出了14种和25种与治疗反应相关的差异蛋白。我们的研究首次从组学角度揭示了NBD的新型蛋白标志物及其与发病机制的关联，上述NBD诊断性与治疗反应性标志物将在更大队列中进行验证。

第三章 神经白塞病脑脊液-血清中枢脱髓鞘标志物研究

神经受累是导致白塞氏综合征（BS）患者预后不良的重要原因。既往尸检结果表明神经白塞病（NBD）中枢神经系统脱髓鞘病变主要发生在炎性血管周围。截止目前缺乏可靠的实验室生物标志物来评估NBD患者的鞘内损伤。本研究旨在探讨髓鞘碱性蛋白（myelin basic protein, MBP），作为一种中枢神经系统髓鞘损伤的检测指标，在NBD患者和疾病对照组中的诊断价值。我们通过ELISA法检测了脑脊液-血清配对样本中MBP的水平，并结合临床常规检测的白蛋白商计算了MBP指数。通过研究发现，NBD患者脑脊液和血清的MBP水平显著高于非炎症性神经系统疾病对照，并且能够以超过90%的特异度将二者有效区分开来。同时脑脊液和血清MBP对急性型和慢性进展型NBD也具有良好的鉴别能力。此外，MBP指数与IgG指数呈正相关。对时序性NBD样本中MBP参数的连续监测发现，血清MBP对疾病复发和药物疗效均具有较好的反应性，而MBP指数则有潜力在临床脱髓鞘症状或影像学表现出现前提示疾病复发。因此，MBP参数对伴有脱髓鞘病变的NBD患者具有较高的诊断价值，并能在影像学或临床诊断之前识别NBD患者中枢神经系统的免疫病理损伤过程。

第四章 神经白塞病中枢免疫球蛋白的合成及其与发病机制的关联研究

神经白塞病（NBD）中枢神经系统免疫反应的存在可由病理组织发现的脱髓鞘和血管周免疫细胞浸润性病灶所证实。本研究旨在探索辅助NBD诊断的免疫球蛋白（Immunoglobulin, Ig）标志物，并通过Ig产生机制揭示NBD潜在病因学特征。我们回顾性纳入了28例具有明确诊断的脑实质受累NBD患者、29例神经精神性红斑狼疮患者、30例中枢神经系统特发性炎性脱髓鞘疾病患者、30例中枢神经系统感染患者、30例脑血管疾病患者和30例非炎症性神经系统疾病患者。采用免疫比浊法检测血清和脑脊液中的Ig，并通过ELISA定量检测髓鞘碱性蛋白水平（MBP）。我们发现：NBD患者的IgA指数比非炎症性神经系统疾病患者高出近两倍，其对二者鉴别诊断较好（AUC=0.8488）；当最佳临界值设置为> 0.6814时，其诊断敏感度和特异度分别为75.00%和90.00%。免疫球蛋白脑脊液水平及其商值（脑脊液与血清中对应Ig的比值）在区分血脑屏障受损和完整的NBD患者时，AUC均超过0.90。此外，聚类分析可将NBD分为两种不同的临床亚组：一种为血脑屏障受损且Ig合成较低；另一种为血脑屏障完整但脑实质部位存在额外的Ig合成。MBP指数与轻链κ指数（r = 0.358）和λ指数（r = 0.575）显著相关（P < 0.001），提示CNS脱髓鞘可能是NBD发病机制中触发鞘内免疫球蛋白过量生成及下游体液免疫反应的关键因素。该研究阐明了IgA指数可作为区分NBD与非炎症性神经系统疾病及中枢神经系统特发性炎性脱髓鞘疾病患者的可靠诊断学指标。中枢神经系统炎症和脱髓鞘事件诱导的过量免疫球蛋白生成可能是NBD潜在的神经免疫学发病机制。

第五章 基于临床实验室参数对血管型白塞病诊断模型的开发与验证

血管型白塞病（Vascular Behçet’s disease, VBD）作为白塞氏综合征（BS）最常见的临床表型，发病率可达40%。炎症诱导的血栓形成在VBD发病中占据主导地位。因此本研究旨在利用机器学习算法平台建立一个用于区分VBD和非VBD患者的炎症-凝血相关诊断性标志物模型。该研究共纳入了338例BS患者，包括123例VBD和215例非VBD患者。我们首先将LassoCV筛选出的26个临床和实验室特征纳入多个机器学习分类器中，以识别用于区分VBD的最佳模型。随后采用Shapley加性解释法识别输入特征对VBD诊断的贡献。并通过逻辑回归分析和列线图筛选与VBD结局相关的危险因素。本研究发现：VBD患者的实验室炎症参数（中性粒细胞百分比、NK细胞、IL-6）、血液学参数（血红蛋白、血红蛋白分布宽度）和凝血参数（活化部分凝血活酶时间、D-二聚体）均升高。我们从XGBoost分类模型中选择贡献度最高的特征应用于BS队列，并使用10折交叉法验证模型诊断性能，发现其诊断效能较强（AUC> 0.90）。利用Shapley加性解释法，我们发现动脉血栓或动脉瘤和深静脉血栓的高发生率、NK细胞计数、血红蛋白分布宽度、活化部分凝血活酶时间和D-二聚体的上调，以及网织红细胞百分比、B细胞计数、红细胞分布宽度、单个红细胞血红蛋白量和TNF-α的下调，构成了对VBD结局的最终解释。通过logistic回归，BS疾病严重程度、血红蛋白、平均红细胞血红蛋白、单个红细胞血红蛋白量、血红蛋白分布宽度、活化部分凝血活酶时间和D-二聚体被识别为BS患者血管受累结局的潜在危险因素。本研究开发了一个基于临床和实验室参数，能够应用于区分VBD的高性能诊断学模型，同时指出炎症和血栓相关实验室指标是VBD潜在的危险因素。

第六章 基于基因公共数据库挖掘白塞氏综合征基因诊断标志物

白塞氏综合征（BS）是一种因免疫细胞异常应答（T细胞亚群失衡及中性粒细胞的过度激活）引起的慢性血管炎，其临床表型具有异质性。HLA-B51是BS发病的关键遗传学易感位点。目前尚缺乏针对免疫细胞的基因表达模式及其在BS遗传学发病机制中的深入研究。本研究中，我们从ArrayExpress下载了E-MTAB-2713数据集，并使用limma包筛选差异表达基因。随后，我们将E-MTAB-2713作为训练集构建了由差异表达基因特征组成的随机森林和神经网络分类模型，并使用GSE17114数据集进行验证。最后，通过单样本基因集富集分析算法评估免疫细胞的浸润情况。我们发现了E-MTAB-2713中BS与对照组差异表达的基因主要富集于病原体触发、淋巴细胞介导、血管生成和糖基化相关的炎症通路。随机森林和神经网络诊断模型中的基因特征，以及在血管生成和糖基化通路中显著富集的基因，能够很好地区分GSE17114中表现为皮肤黏膜、眼部和静脉血栓受累的BS临床亚型。此外，与健康对照组相比，中性粒细胞、T细胞、NK细胞和浆细胞样树突状细胞构成了BS独特的免疫细胞浸润图谱。我们的研究结果表明，CD14+单核细胞中EPHX1、PKP2、EIF4B和HORMAD1的表达，以及CD16+中性粒细胞中CSTF3和TCEANC2的表达，可作为区分BS表型的重要基因学特征。同时，血管生成相关通路基因（如ATP2B4、MYOF和NRP1）和糖基化相关通路基因（如GXYLT1、ENG、CD69、GAA、SIGLEC7、SIGLEC9和SIGLEC16）也可能作为识别BS临床表型的诊断学标志物。

Behçet’s Syndrome (BS) is a variable vasculitis with heterogenous clinical features, primarily characterized by the classic triad of oral ulcers, genital ulcers, and uveitis. When affecting arteries, veins, the gastrointestinal tract, or the nervous system, it can cause severe systemic damage and even death. Currently, the diagnostic criteria for BS are based on clinical symptoms and mucocutaneous manifestations, yet there is still a lack of effective laboratory diagnostic markers. The dynamic clinical presentation of disease activity (relapse and remission phases) and the complex overlap of clinical phenotypes pose significant challenges in identifying biomarkers for BS. Therefore, there is an urgent need to discover biomarkers related to differential diagnosis, phenotyping, and disease progression (activity and severity) for early diagnosis, treatment response monitoring, and disease course assessment. This study focuses on screening and validating BS-specific biomarkers by investigating plasma autoantibodies, cerebrospinal fluid-blood paired proteomics, routine laboratory indicators, and genomics. We aim to construct biomarker models for BS diagnosis and clinical phenotypes, as well as to explore the association between specific markers and the pathogenesis of BS. The main research content of this thesis is divided into the following six chapters:

Chapter I Autoantibody profiling of Behcet's syndrome

There lack autoantibody biomarkers for Behçet’s Syndrome that reflect its clinically heterogeneous phenotypes and disease course, which hinders standard diagnosis and precise treatment. Therefore, we utilized a high-throughput protein microarray platform to identify broad-spectrum and phenotype-specific autoantibodies in BS. In this study, a total of 840 plasma samples including 495 BS patients, 227 disease controls, and 118 healthy controls were collected. An established strategy involving large-scale microarray primary screening, focused microarray validation, and ELISA verification was employed to identify BS autoantibody biomarkers. Machine learning algorithm was used to optimize and construct autoantibody panels for BS diagnosis. Through differential analysis, pathway enrichment, and clustering methods, we elucidated the importance of phenotype-specific and severity-related autoantibodies with regard to the immunopathological mechanisms of BS organ/tissue damage. Additionally, immunohistochemical staining was performed on BS tissue sections to detect the localization of targeted antigens by autoantibodies. Mendelian randomization was used to validate the causal relationship between the expression of target antigen-related gene polymorphisms and BS clinical manifestations. First, primary screening using the large-scale microarray revealed that differentially expressed antigens between BS and healthy controls were significantly enriched in pathways such as platelet degranulation, chaperone-mediated autophagy, and neutrophil activation. Second, validated by focused microarray, we constructed a machine learning model comprising 19 candidate autoantibodies to well-differentiate BS from controls, the AUC of which was 0.82. Furthermore, anti-HBS1L antibody was verified as a "broad-spectrum" marker for BS diagnosis, while anti-PPP1R13L antibody showed potency for assessing disease severity. Additionally, anti-p017 antibody was identified as a phenotype-specific antibody for gastrointestinal phenotype of BS. Finally, immunohistochemical results revealed that CCDC140, primarily expressed in the cytoplasmic matrix and internal mucus of colon tissues among BS patients, is a major target of autoantibodies leading to gastrointestinal lesions, which could be also used to distinguish the gastrointestinal phenotype of BS from inflammatory bowel disease patients. In this study, we also delineated an underlying interim status of BS transitioning from autoinflammatory to autoimmune onset, and also provided potential autoantibody candidates for clinical diagnosis of BS.

Chapter II Proteomic study of paired cerebrospinal fluid and blood samples in Neuro-Behçet’s Disease

Neuro-Behçet’s disease (NBD) is a severe neurological phenotype of Behçet’s Syndrome (BS), which is relatively rare in clinical practice and is mainly classified into parenchymal and non-parenchymal subtypes. Pathologically, perivascular inflammation accompanied by immune cell infiltration around the central nervous system (CNS) and elevated cerebrospinal fluid (CSF) IL-6 levels are characteristic neuroinflammatory manifestations among NBD patients. To date, few studies have revealed biomarkers that are specific to the disease course or pathogenesis of NBD from a high-throughput proteomic perspective. In this study, we employed data-independent acquisition mass spectrometry to conduct comprehensive proteomic screening of paired CSF and blood samples from NBD patients and disease controls. We preliminarily analyzed the differentially expressed proteins and their associated pathogenic biological pathways at both CSF and plasma levels. From the aspect of CSF compartment, we identified 117 differentially expressed proteins primarily involved in complement activation and humoral immune response pathways. By comparing with the Human Protein Atlas database, we found that these proteins are mainly expressed in the cerebral cortex, cerebellum, and caudate nucleus, and dominantly regulate the reactive oxygen species response pathways. As for the plasma level, 79 differentially expressed proteins were significantly upregulated in NBD and were associated with acute inflammation, coagulation, axonal development, and humoral immune pathways. Among these, proteins markers primarily localized in the CNS exert their biological functions by regulating endoplasmic reticulum stress responses. In terms of diagnostic value, we focused on 14 proteins that were differentially expressed in both CSF and plasma of NBD patients comparing with control subjects, aiming to identify the "non-invasive" surrogate biomarkers. We preliminarily found that the combined detection of VSIG4 protein in CSF and plasma could well distinguish NBD patients from controls (AUC= 0.824). In the differential expression analysis of longitudinal NBD samples, we respectively identified 14 and 25 differentially expressed proteins in CSF and plasma, that were associated with beneficial therapeutic response. Our study first reveals the novel protein biomarkers and their associated pathogenic mechanisms in NBD from a proteomic perspective. These diagnostic and therapeutic responsiveness markers for NBD will be further validated in larger cohorts.

Chapter III Cerebrospinal fluid and serum biomarkers for delineating central demyelination in Neuro-Behçet’s Disease

Neuro-Behçet’s disease (NBD) is a main cause of poor prognosis among patients with Behçet’s syndrome (BS). Previous autopsy study has indicated that demyelinating lesions in the central nervous system (CNS) of NBD patients primarily occur around the inflammatory blood vessels. To date, there is a lack of reliable laboratory biomarkers to assess intrathecal damage in NBD patients. This study was to explore the diagnostic value of myelin basic protein (MBP), a dimension for CNS myelin damage, in NBD patients and disease control groups. We detected MBP levels in paired cerebrospinal fluid (CSF) and serum samples using ELISA and calculated the MBP index by using quotient of albumin from routine measurements on clinic. We found that CSF and serum MBP levels in NBD patients were significantly higher than those with non-inflammatory neurological diseases, which could effectively differentiate them with a specificity above 90%. Additionally, CSF and serum MBP demonstrated good discriminative ability between acute and chronic progressive NBD. Furthermore, MBP index showed a positive correlation with the IgG index. Serial monitoring of MBP in longitudinal NBD cohort revealed that serum MBP is responsive to disease recurrence and therapeutic benefits, while the MBP index has the potency to indicate disease relapse before clinical symptoms or imaging manifestations on demyelinating events. Therefore, MBP parameters held a high diagnostic value for NBD patients with demyelinating lesions, which could identify the immunopathological deteriorations in the CNS prior to imaging or clinical manifestations.

Chapter IV Association between central immunoglobulin synthesis and their immunopathogenesis in Neuro-Behçet’s disease

The existence of neuro-immune responses in the central nervous system (CNS) of Neuro-Behçet’s disease (NBD) is supported by pathological findings of demyelination together with perivascular immune cell infiltration. This study aims to explore the immunoglobulin (Ig) for the auxiliary diagnosis and their pathological contributions to NBD. We retrospectively enrolled 28 patients with confirmed parenchymal NBD, 29 patients with neuropsychiatric systemic lupus erythematosus, 30 patients with CNS idiopathic inflammatory demyelinating diseases (CNS-IIDD), 30 patients with CNS infections, 30 patients with cerebrovascular diseases, and 30 patients with non-inflammatory neurological diseases (NIND). Ig levels in serum and cerebrospinal fluid (CSF) were measured using immunonephelometry, and myelin basic protein (MBP) levels were quantified using enzyme-linked immunosorbent assay (ELISA). Our study found that with an AUC of 0.8488, the IgA index in NBD patients was nearly twice as high as those with NIND. When the optimal cutoff value was set at > 0.6814, the sensitivity and specificity of IgA index in differentiating NBD and NIND were 75.00% and 90.00%, respectively. Immunoglobulin levels in CSF and their quotients (ratio of CSF Ig/Serum Ig) could accurately differentiate NBD patients with damaged blood-brain barrier (BBB) versus those with intact ones (AUC>0.90). Furthermore, cluster analysis divided NBD patients into two distinct clinical subgroups: ones holding BBB damage but lower Ig synthesis, and the others with intact BBB but overproduction of Ig synthesis in the parenchymal sites. Then, the compact correlations showed between MBP index and kappa (κ) index (r = 0.358) and lambda (λ) index (r = 0.575) (P < 0.001), suggested that CNS demyelination may be a key factor of triggering excessive intrathecal Ig production and downstream humoral immune responses in NBD pathogenesis. This study demonstrates that the IgA index can serve as a reliable diagnostic marker for distinguishing NBD from NIND and CNS-IIDD. As a consequence of neuroinflammation and demyelination, excessive Ig production may uncover a potential neuro-immunological mechanism underlying among NBD patients.

Chapter V Development and validation of a diagnostic panel for vascular Behcet's disease based on clinical and laboratory parameters

Vascular Behçet’s disease (VBD), as the most common clinical phenotype of Behçet’s syndrome (BS), holding a prevalence rate above 40%. Inflammation-induced thrombosis plays an indispensable role in the pathogenesis of VBD. Therefore, this study aims to establish an inflammation-thrombosis diagnostic panel for distinguishing VBD from non-VBD patients using a machine learning algorithm engined platform. A total of 338 BS patients were enrolled, including 123 VBD and 215 non-VBD patients. First, 26 clinical and laboratory features selected by LassoCV were incorporated into multiple classifiers to identify the optimal model for differentiating VBD. The Shapley Additive Explanations (SHAP) method was then used to interpret the contribution of input features to VBD diagnosis. Logistic regression analysis and nomogram were conducted to screen the risk factors associated with VBD outcomes. We found that laboratory inflammatory parameters (neutrophil percentage, NK cells, IL-6), hematological parameters [hemoglobin, hemoglobin distribution width (HDW)], and coagulation parameters [activated partial thromboplastin time (APTT), D-dimer] were elevated in VBD patients. Then, the top features from the XGBoost classification model were applied to the BS validation cohort. Using these features, the diagnostic performance of the model was trained with 10-fold cross-validation method, which achieved an excellent diagnostic value (AUC>0.90). By SHAP method, we identified that a higher incidence of arterial thrombosis or aneurysms and deep vein thrombosis, upregulated NK cell count, HDW, APTT, and D-dimer, as well as downregulated reticulocyte%, B cell count, red blood cell distribution width, cellular hemoglobin (CH), and TNF-α, collectively provided the final explanation for VBD outcomes. Through logistic regression, BS disease severity, hemoglobin, mean corpuscular hemoglobin, CH, HDW, APTT, and D-dimer were identified as the potential risk factors for vascular involvement among BS patients. To summarize, this study developed a high-performance diagnostic panel based on clinical and laboratory parameters for distinguishing VBD, while also highlighted that inflammation and thrombosis risk factors as potential contributors to VBD.

Chapter VI Preliminary exploration of the genetic markers for Behcet's syndrome based on public gene datasets

Behçet’s syndrome (BS) is a chronic vasculitis with heterogeneous clinical phenotypes, which is characterized by immune cell abnormalities (imbalance in T-cell subsets and overactivation of neutrophils). Genetic susceptibility, particularly the HLA-B51 gene locus, is a key factor in BS pathogenesis. Currently, there is a lack of comprehensive research on gene expression patterns among immune cells and their associations with BS etiology. Herein, we downloaded the E-MTAB-2713 dataset from ArrayExpress and used the R limma package to screen for differentially expressed genes. Using E-MTAB-2713 as the training set, we constructed random forest and neural network classification models based on differential gene expression features and validated them using the GSE17114 dataset. Immune cell infiltration was assessed using single-sample gene set enrichment analysis. Differentially expressed genes between BS and control groups in E-MTAB-2713 were primarily enriched in pathogen-triggered, lymphocyte-mediated, angiogenesis-associated, and glycosylation-related inflammatory pathways. Genetic features from the random forest and neural network diagnostic models, along with genes enriched in angiogenesis and glycosylation pathways, could effectively distinguish BS clinical subtypes in GSE17114, including mucocutaneous, ocular, and venous thrombosis involvement. Additionally, compared to healthy controls, neutrophils, T cells, NK cells, and plasmacytoid dendritic cells constituted a unique immune cell infiltration profile in BS. Our findings revealed that the expression of EPHX1, PKP2, EIF4B, and HORMAD1 in CD14+ monocytes, as well as CSTF3 and TCEANC2 in CD16+ neutrophils, can serve as genetic signatures for distinguishing BS phenotypes. Angiogenesis-related (ATP2B4, MYOF, and NRP1) and glycosylation-related pathway genes (GXYLT1, ENG, CD69, GAA, SIGLEC7, SIGLEC9, and SIGLEC16) could also be considered as diagnostic markers for identifying BS phenotypes.