【作者】 李岩；

【导师】 张志民；

【作者基本信息】 吉林大学 ， 口腔医学硕士（专业学位）， 2025， 硕士

【摘要】 研究背景:头颈部鳞状细胞癌(Head and Neck Squamous Cell Carcinoma,HNSCC)被认为是全球第六大常见癌症。尽管针对HNSCC的手术、放疗、化疗以及免疫治疗等方法被广泛使用,但局部复发和远处转移仍然是影响患者生存率的重要难题。目前HNSCC患者的5年生存率约50%。随着信息科学和分子生物学的进步,尤其是大量在基因层面的研究让我们对肿瘤的发生发展有了更深层的认知。以往的研究指出研究HNSCC的关键预后基因,并结合早期筛查手段,有望实现更精确的风险评估。乳酰化是一种蛋白质翻译后修饰,它能够通过转录调控影响癌症的生长、转移、侵袭及肿瘤免疫逃避。因此,乳酰化相关基因有可能成为HNSCC新的潜在预后标志物。研究目的:本研究旨在通过生物信息学和机器学习算法构建基于乳酰化相关基因的HNSCC预后模型,以研究HNSCC患者的预后和临床治疗效果。研究方法:1.从癌症基因组图谱(The Cancer Genome Atlas,TCGA)数据库和基因表达综合(Gene Expression Omnibus Database,GEO)数据库中下载并整理了HNSCC患者的RNA测序数据及临床数据。乳酰化相关基因(Lactylation-Related Genes,LRGs)是从分子特征数据库(Molecular Signatures Database,MSig DB)中获取和筛选的。然后通过R软件包“limma”筛选差异表达基因(Differentially Expressed Genes,DEGs),使用STRING数据库和Cytoscape软件构建蛋白质-蛋白质相互作用(Protein-Protein Interaction,PPI)网络。2.通过单因素Cox回归分析、随机生存森林(Random Survival Forest,RSF)分析及多因素Cox回归分析,筛选出3个与HNSCC预后显著相关的乳酰化相关基因。基于这3个核心基因构建乳酰化相关预后模型(Lactylation-Related Prognostic Model,LRPM)。根据风险评分将患者分为高风险组和低风险组。使用主成分分析(Principal Component Analysis,PCA)、Kaplan-Meier生存分析和受试者工作特征曲线(Receiver Operating Characteristic Curves,ROC)验证模型的预测效果。3.通过CIBERSORT、免疫细胞丰度识别器(Immune Cell Abundance Identifier,Immu Cell AI)、ESTIMATE、TIDE和免疫表观评分(Immunophenoscore,IPS)评估两组患者的免疫细胞浸润模式及对免疫检查点抑制剂(Immune Checkpoint Inhibitors,ICIs)的反应,并通过onco Predict和Cell Miner数据库筛选出潜在的治疗药物。利用人类蛋白质图谱(Human Protein Atlas,HPA)和TCGA数据库进行多组学验证(包括蛋白表达、突变频率及甲基化状态分析)。4.临床收集HNSCC患者的肿瘤组织及其癌旁正常组织样本,同时体外培养人舌鳞癌细胞系CAL-27和人口腔上皮细胞系HOEC,通过实时荧光定量多聚核苷酸链式反应(Quantitative Real-time polymerase chain reaction,RT-qPCR)验证肿瘤组织与癌旁正常组织中、CAL-27与HOEC中核心基因表达水平的差异,对预后模型进行进一步的验证。研究结果:1.整理TCGA和GEO数据库中HNSCC患者的RNA测序数据及临床数据和MSig DB数据库中858个LRGs。使用“limma”R包筛选出17237个DEGs,其中包括104个与乳酰化相关的差异表达基因(Differentially Expressed Lactylation-Related Genes,DELRGs)。随后通过STRING数据库和Cytoscape软件构建了DELRGs的PPI网络,揭示了这些基因的潜在功能关联。2.通过单因素Cox回归分析、RSF分析以及多因素Cox回归分析,筛选出3个核心基因(PABPN1、STC2和SATB1)并构建LRPM。根据风险评分公式计算风险评分,将患者分为高风险组和低风险组。Kaplan-Meier生存分析和ROC曲线验证显示,高风险组的总生存概率显著低于低风险组。PCA进一步证实了模型的分组效果。3.CIBERSORT、Immu Cell AI和ESTIMATE算法结果显示低风险组中免疫激活相关细胞显著富集,而高风险组则以免疫抑制细胞为主。TIDE和IPS分析显示,低风险组对ICIs的反应高于高风险组。利用onco Predict和Cell Miner数据库筛选出潜在的治疗药物,为不同风险组提供了个性化治疗方案。通过HPA和TCGA数据库对DELRGs进行多组学验证,证实了STC2和SATB1的高表达与较差预后相关,而PABPN1的高表达则与较好预后相关。4.RT-qPCR验证了核心基因在肿瘤组织与癌旁正常组织中的表达存在显著差异,在CAL-27舌鳞癌细胞系和HOEC口腔上皮细胞系中核心基因表达水平亦存在差异。研究结论:本研究发现乳酰化相关基因特征可作为预测头颈部鳞状细胞癌患者预后的潜在生物标志物,基于生物信息学和机器学习算法构建的乳酰化相关预后模型可用于分析其肿瘤免疫微环境及临床治疗效果,为新型的个体化治疗策略提供新的思路。更多还原

【Abstract】 Backgroud:Head and neck squamous cell carcinoma(HNSCC)is considered the sixth most common cancer worldwide.Although surgery,radiotherapy,chemotherapy,and immunotherapy are widely used for HNSCC,local recurrence and distant metastasis are still important challenges affecting patient survival.Currently,the 5-year survival rate of HNSCC patients is about 50%.With the advancement of information science and molecular biology,especially a large number of studies at the genetic level have given us a deeper knowledge of tumor development.Previous studies say that studying key prognostic genes in head and neck squamous cell carcinoma and combining them with early screening tools is expected to lead to a more accurate risk assessment.Lactylation is a post-translational modification of proteins that can affect cancer growth,metastasis,invasion and tumor immune evasion through transcriptional regulation.Therefore,lactylation-related genes may become new potential prognostic markers for HNSCC.Objective:The aim of this study was to construct a prognostic model of HNSCC based on lactylation-related genes by bioinformatics and machine learning algorithms in order to study the prognosis and clinical outcomes of HNSCC patients.Methods:1.RNA sequencing data and clinical data of HNSCC patients were downloaded and organized from The Cancer Genome Atlas(TCGA)database and Gene Expression Omnibus Database(GEO)database.Lactylation-Related Genes(LRGs)were obtained and screened from the Molecular Signatures Database(MSig DB).Differentially Expressed Genes(DEGs)were then screened by the“limma”R package,and protein-protein interactions(PPIs)were constructed using the STRING database and Cytoscape software.2.Three lactylation-related genes significantly associated with the prognosis of HNSCC were screened by one-way Cox regression analysis,Random Survival Forest(RSF)analysis and multifactorial Cox regression analysis.Lactylation-Related Prognostic Model(LRPM)was constructed based on these three core genes.Patients were categorized into high-risk and low-risk groups based on risk scores.The predictive effect of the model was verified using Principal Component Analysis(PCA),Kaplan-Meier survival analysis and Receiver Operating Characteristic Curves(ROC).3.The pattern of immune cell infiltration and response to immune checkpoint inhibitors(ICIs)were assessed by CIBERSORT,Immune Cell Abundance Identifier(Immu Cell AI),ESTIMATE,TIDE,and Immunophenoscore(IPS)in the two groups of patients,and potential therapeutic agents were screened by onco Predict and Cell Miner database.Inhibitors(ICIs),and potential therapeutic agents were screened by onco Predict and Cell Miner databases.Multi-omics validation(including protein expression,mutation frequency and methylation status analysis)was performed using Human Protein Atlas(HPA)and TCGA databases.4.Tumors and their paracancerous normal tissue samples from HNSCC patients were collected clinically,while the human tongue squamous carcinoma cell line CAL-27 and the human oral epithelial cell line HOEC were cultured in vitro to validate the tumor and normal tissues in tumors and paracancerous normal tissues by real-time fluorescence quantitative polymerase chain reaction(RT-qPCR)to verify the differences in the expression levels of three key genes in CAL-27 and HOEC between tumor and normal tissues for further validation of the prognostic model.Results:1.The RNA sequencing data and clinical data of HNSCC patients in TCGA and GEO databases and 858 LRGs in MSig DB database were collated,and 17,237 DEGs were screened using the“limma”R package,including 104 Differentially Expressed Lactylation-Related Genes(DELRGs),which were associated with lactylation.Subsequently,the PPI network of DELRGs was constructed by STRING database and Cytoscape software,which revealed the potential functional associations of these genes.2.Three core genes(PABPN1,STC2,and SATB1)were screened and LRPM was constructed by one-way Cox regression analysis,RSF analysis,and multifactorial Cox regression analysis.The patients were divided into high-risk and low-risk groups by calculating the risk score according to the risk score formula.The Kaplan-Meier survival analysis and the validation of ROC curves showed that the total survival probability was significantly lower than that of the low-risk group.PCA further confirmed the model’s grouping effect.3.The results of CIBERSORT,Immu Cell AI and ESTIMATE algorithms showed a significant enrichment of anti-tumor immune cells in the low-risk group,whereas the high-risk group was dominated by immune-suppressive cells.TIDE and IPS analyses showed that the low-risk group responded to ICIs more than the high-risk group.Seventy potential therapeutic agents were screened using onco Predict and Cell Miner databases,which provided personalized treatment options for different risk groups.Multi-omics validation of DELRGs by HPA and TCGA databases confirmed that high expression of STC2 and SATB1 was associated with poorer prognosis,while high expression of PABPN1 was associated with better prognosis.4.RT-qPCR verified that there were significant differences in the expression of the core genes in tumor tissues and normal tissues adjacent to the cancer,as well as differences in the expression levels of the core genes in the CAL-27 tongue squamous carcinoma cell line and the HOEC oral epithelial cell line.Conclusions:In this study,we found that lactylation-related gene features can be used as potential biomarkers for predicting the prognosis of patients with head and neck squamous cell carcinoma,and LRPM constructed based on bioinformatics and machine learning algorithms can be used to analyze their tumor immune microenvironments and clinical therapeutic efficacies,which will provide new ideas for novel and individualized therapeutic strategies.更多还原