Table 3_Innate immune cell barrier-related genes inform precision prognosis in pancreatic cancer.xlsx

IntroductionPancreatic cancer (PC) remains a lethal malignancy with limited treatment options. The role of innate immune cell barrier-related genes in PC prognosis is poorly defined. This study aimed to identify prognostic biomarkers, develop a predictive model, and uncover novel targets for personalized therapy.
MethodsInnate immune cell barrier-related genes were curated from KEGG, ImmPort, MSigDB, and InnateDB. Differential expression analysis was performed using TCGA and GTEx datasets. Univariate Cox regression identified survival-associated genes. Prognostic modeling of PC was developed using 14 machine learning algorithms, with performance validated through long-term survival metrics, functional enrichment, immune infiltration analysis, and drug sensitivity profiling. Core genes were prioritized via the "mime1" package, and single-cell RNA sequencing (scRNA-seq) data explored UBASH3B’s functional role.
Results352 differentially expressed genes of Innate immune cell barrier-related were identified, with NK cell pathways linked to PC immunity. Univariate Cox analysis revealed 8 protective and 84 risk genes. The RSF model (trained on risk genes) showed strong 3- and 5-year survival prediction. High-risk patients exhibited elevated tumor mutation burden (TMB), reduced NK/CD8+ T cell infiltration, and resistance to Erlotinib/Oxaliplatin but sensitivity to 5-Fluorouracil. Five key genes (ITGB6, COL17A1, MMP28, DIAPH3, UBASH3B) were highlighted. UBASH3B, a novel marker, correlated negatively with NK cell activation and mediated immune signaling and drug resistance.
DiscussionThis study established the CDRG-RSF model, a robust prognostic tool leveraging innate immune genes. UBASH3B’s dual role in immune suppression and drug resistance highlights its potential for stratifying PC patients into tailored treatment groups. The findings underscore the importance of integrating machine learning with immune profiling to advance precision oncology for PC.

引言：胰腺癌（PC）仍是一种治疗手段有限的致死性恶性肿瘤，目前学界对先天免疫细胞屏障相关基因在胰腺癌预后中的作用尚不清楚。本研究旨在识别预后生物标志物、构建预测模型，并发掘个性化治疗的新型靶点。
方法：从京都基因与基因组百科全书（KEGG）、免疫端口数据库（ImmPort）、分子特征数据库（MSigDB）及先天免疫数据库（InnateDB）中筛选先天免疫细胞屏障相关基因。采用癌症基因组图谱（TCGA）与基因型组织表达（GTEx）数据集进行差异表达分析。通过单变量Cox回归分析识别与生存相关的基因。本研究使用14种机器学习算法构建胰腺癌预后模型，并通过长期生存指标、功能富集分析、免疫浸润分析以及药物敏感性谱验证模型性能。通过"mime1"R包筛选核心基因，并借助单细胞RNA测序（single-cell RNA sequencing，scRNA-seq）数据探究UBASH3B的功能作用。
结果：共鉴定出352个先天免疫细胞屏障相关差异表达基因，其中自然杀伤细胞（natural killer cell，NK细胞）通路与胰腺癌免疫密切相关。单变量Cox分析筛选出8个保护基因与84个风险基因。基于风险基因训练的随机生存森林（Random Survival Forest，RSF）模型对3年及5年生存率具有良好的预测性能。高危患者的肿瘤突变负荷（tumor mutation burden，TMB）升高，NK细胞/CD8+ T细胞浸润减少，且对厄洛替尼（Erlotinib）、奥沙利铂（Oxaliplatin）耐药，但对5-氟尿嘧啶（5-Fluorouracil）敏感。最终筛选出5个关键基因：ITGB6、COL17A1、MMP28、DIAPH3及UBASH3B。新型标志物UBASH3B与NK细胞活化呈负相关，可介导免疫信号通路与药物耐药。
讨论：本研究构建的CDRG-RSF模型是一种基于先天免疫基因的稳健预后工具。UBASH3B在免疫抑制与药物耐药中的双重作用，凸显了其可用于将胰腺癌患者分层至个体化治疗组的潜力。本研究结果强调了将机器学习与免疫特征分析相结合，对推进胰腺癌精准肿瘤学发展的重要意义。