通过机器学习和多组学分析鉴定GADD45A为脓毒症中潜在的R-loop调节因子
该研究整合4个GEO芯片数据集,经批次校正和标准化后行差异表达分析和WGCNA,取交集获得重叠R-loop相关基因(RLRGs),再用机器学习优选枢纽基因并构建诊断模型,在外部队列验证。结果鉴定出EXOSC4、GADD45A、HMGB2、LDHA、PRKDC五个重叠RLRGs,均被最优机器学习模型保留;五基因列线图校准良好且具临床效用。单细胞分析显示其在中性粒细胞、单核细胞和T细胞中呈细胞类型特异性表达;仅GADD45A与生存不良显著相关。盲肠结扎穿孔脓毒症小鼠模型中GADD45A上调;HL-60细胞中沉默GADD45A可减少LPS诱导的炎性细胞因子表达、R-loop积累和DNA损伤。
要点
- 在4个GEO数据集整合分析中,EXOSC4、GADD45A、HMGB2、LDHA、PRKDC五个重叠R-loop相关基因均被最优机器学习模型保留,支持其稳健性。
- 基于上述五基因构建的列线图诊断模型校准良好且具临床效用,并在外部队列中验证。
- 五个基因中仅GADD45A与脓毒症患者生存不良显著相关,提示其预后价值。
- 盲肠结扎穿孔诱导的脓毒症小鼠模型中GADD45A表达上调。
- HL-60细胞中沉默GADD45A可减少LPS诱导的炎性细胞因子表达、R-loop积累和DNA损伤。
适用人群与场景
- 研究基于4个Gene Expression Omnibus(GEO)芯片数据集整合分析,纳入脓毒症相关样本,具体纳入/排除标准原文未提及。
- 外部验证队列人群及盲肠结扎穿孔脓毒症小鼠模型、HL-60细胞实验条件原文未详细提及。
诊断标准与关键阈值
- 原文未提及脓毒症的临床诊断标准或评分界值。
处置流程与药物剂量
- 原文未提及针对脓毒症的临床处置流程或药物剂量。
红旗征与不能漏的情况
- 原文未提及需要立刻处理或不能漏诊的红旗征。
研究设计与关键结果
- 研究类型为基于公共芯片数据集的多组学与机器学习分析,整合4个GEO数据集,具体样本量与中心数原文未提及。
- 通过差异表达分析和WGCNA取交集获得5个重叠R-loop相关基因:EXOSC4、GADD45A、HMGB2、LDHA、PRKDC,均被最优机器学习模型保留。
- 单细胞RNA测序显示五个基因在中性粒细胞、单核细胞和T细胞中呈细胞类型特异性表达。
- 五个基因中仅GADD45A与生存不良显著相关;盲肠结扎穿孔脓毒症小鼠模型中GADD45A上调。
- HL-60细胞中沉默GADD45A可减少LPS诱导的炎性细胞因子表达、R-loop积累和DNA损伤。
证据类型、等级与局限
- 证据类型为基于公共数据库的生物信息学分析结合体内外实验验证的转化研究,非临床RCT或队列研究。
- 局限:基于公共芯片数据集整合分析,样本来源和临床特征受原始数据集限制;机制实验仅在HL-60细胞和CLP小鼠模型中进行,尚未在临床样本中验证GADD45A作为诊断/预后标志物的阈值和决策价值。
对床旁实践的改变
- 原文结果尚不能直接改变急诊/ICU对脓毒症的诊断或治疗流程;GADD45A作为诊断和预后生物标志物仍需前瞻性临床验证。
- 适用与不适用人群、常见陷阱原文未提及。
指南或共识出处
- 原文未提及
以上提炼依据论文摘要(全文未开放获取,或许可不允许翻译转载);全文请点「阅读原文」。
英文原文(摘要)· 点击展开
BACKGROUND: Sepsis is a life-threatening syndrome characterized by dysregulated immune responses, while the molecular basis of immune dysfunction remains unclear. Increasing evidence suggests that R-loop accumulation and DNA damage may contribute to immune disorders, but their roles in sepsis have not been systematically investigated.
METHODS: Four Gene Expression Omnibus microarray datasets were integrated after batch-effect correction and normalization. Differential expression analysis and weighted gene co-expression network analysis were performed to identify sepsis-related genes. Overlapping R-loop-related genes (RLRGs) were obtained by intersecting differentially expressed genes, key module genes, and curated RLRGs. Machine learning was then applied to prioritize hub genes and construct an optimal diagnostic model, which was validated in external cohorts. A nomogram, single-cell RNA sequencing analysis, survival analysis, and in vivo / in vitro experiments were further used for validation.
RESULTS: Five overlapping RLRGs ( EXOSC4, GADD45A, HMGB2, LDHA , and PRKDC ) were identified, and all were consistently retained by top-performing machine learning models, supporting their robustness. The five-gene nomogram showed good calibration and clinical utility. Single-cell analysis revealed cell-type-specific expression across neutrophils, monocytes, and T cells. Among the five genes, only GADD45A was significantly associated with poor survival. Experimental validation showed that GADD45A was upregulated in a cecal ligation and puncture-induced sepsis mouse model. In HL-60 cells, GADD45A silencing reduced lipopolysaccharide-induced inflammatory cytokine expression, R-loop accumulation, and DNA damage.
CONCLUSIONS: GADD45A is a promising biomarker for the diagnosis and prognosis of sepsis and may promote disease progression by enhancing R-loop accumulation and DNA damage.
参考来源
- Chen M, Kang JX, He SL, et al. Identifying GADD45A as a Potential R-Loop Regulator in Sepsis Via Machine-Learning and Multi-Omics Analysis. Shock. 2026. PMID: 42560974. doi:10.1097/SHK.0000000000002902
供执业医师抢救与诊疗参考,不替代临床判断;剂量请结合体重、肝肾功能和本院药品规格复核,以指南原文为准。