Table 4_Multi-omics integration reveals gut microbiota dysbiosis and metabolic alterations of cerebrospinal fluid in children with epilepsy.xlsx

IntroductionEpilepsy is a complex neurological disorder with an unclear pathogenesis. Emerging evidence suggests that gut microbiota dysbiosis and cerebrospinal fluid (CSF) metabolic alterations play a critical role in epilepsy progression through the gut–brain axis. This study aimed to characterize microbial and metabolic disturbances in pediatric epilepsy and identify potential diagnostic biomarkers through integrative multi-omics analysis of matched fecal and CSF samples.
MethodsIn this study, we conducted 16S rRNA gene sequencing on fecal samples from a total of 50 participants including 17 common epilepsy (CEP) patients, 23 refractory epilepsy (REP) patients, and 10 non-epilepsy (NEP) patients, along with untargeted metabolomic analysis on 24 paired CSF samples from REP and NEP groups. Multi-omics integration and a random forest model were applied to assess diagnostic performance, identifying microbial and metabolite signatures associated with epilepsy.
ResultsChildren with epilepsy (REP and CEP) exhibited distinct gut microbiota dysbiosis. Specifically, multivariable association modeling using MaAsLin 3 identified 13 discriminatory microbial taxa, with Clostridiales and Clostridiaceae ranking as the most enriched in REP. Functional predictions revealed significant differences in metabolic pathway, alongside disrupted ecological characteristics among epilepsy groups. In addition, CSF metabolomics analysis further revealed key metabolic shifts between REP and NEP, with notable alterations in alpha-Ketoisocaproic acid, alpha-Ketoisovaleric acid, and acetyl-L-carnitine, reflecting distinct metabolic reprogramming in epilepsy. Moreover, correlation analysis revealed strong microbiota-metabolite associations, reinforcing the involvement of the gut-brain axis in epileptogenesis. Independent random forest-based diagnostic models using microbial genera (AUC = 0.913, accuracy = 0.818) or metabolites (AUC = 0.875, accuracy = 0.833) demonstrated high classification accuracy in distinguishing REP from NEP. Notably, the integrated microbiota-metabolite classification model exhibited superior diagnostic performance in REP and NEP groups (AUC = 0.953, accuracy = 0.875), significantly surpassing individual models and highlighting the potential of multi-omics integration for epilepsy diagnostics.
ConclusionThese findings reveal concurrent gut microbiota dysbiosis and CSF metabolic disturbances in epilepsy, underscoring their interrelated roles in epileptogenesis and reinforcing our understanding of microbiome-metabolome crosstalk. The integrated multi-omics model demonstrated superior diagnostic performance, emphasizing its potential for precision biomarker discovery and clinical application in epilepsy stratification and intervention.

引言
癫痫是一种发病机制尚未明确的复杂神经系统疾病。越来越多的研究证据表明，肠道菌群失调（gut microbiota dysbiosis）与脑脊液（cerebrospinal fluid, CSF）代谢异常通过肠-脑轴（gut–brain axis）在癫痫进展过程中发挥关键作用。本研究旨在通过对匹配的粪便与脑脊液样本进行整合多组学分析（integrative multi-omics analysis），明确儿童癫痫（pediatric epilepsy）患者的微生物与代谢紊乱特征，并筛选潜在的诊断生物标志物（diagnostic biomarkers）。
方法
本研究共纳入50名受试者，其中包括17例普通型癫痫（common epilepsy, CEP）患者、23例难治性癫痫（refractory epilepsy, REP）患者以及10例非癫痫（non-epilepsy, NEP）对照者，对所有受试者的粪便样本进行16S rRNA基因测序（16S rRNA gene sequencing）；同时对难治性癫痫与非癫痫对照组的24例配对脑脊液样本开展非靶向代谢组学分析（untargeted metabolomic analysis）。本研究采用多组学整合策略结合随机森林模型（random forest model）评估诊断效能，筛选与癫痫相关的微生物与代谢物特征谱。
结果
癫痫儿童（包括难治性癫痫与普通型癫痫患者）表现出显著的肠道菌群失调。具体而言，采用MaAsLin 3进行多变量关联建模，共筛选出13个具有鉴别能力的微生物分类群，其中梭菌目（Clostridiales）与梭菌科（Clostridiaceae）在难治性癫痫患者中富集程度最高。功能预测分析显示，癫痫组间的代谢通路存在显著差异，同时生态特征也出现紊乱。此外，脑脊液代谢组学分析进一步揭示了难治性癫痫与非癫痫对照组之间的关键代谢变化，其中α-酮异己酸（alpha-Ketoisocaproic acid）、α-酮异戊酸（alpha-Ketoisovaleric acid）与乙酰-L-肉碱（acetyl-L-carnitine）出现显著改变，反映出癫痫患者存在独特的代谢重编程现象。相关性分析还发现微生物组与代谢组之间存在较强的关联，进一步证实了肠-脑轴参与癫痫发生过程。基于微生物属水平数据构建的独立随机森林诊断模型，其曲线下面积（Area Under Curve, AUC）为0.913，准确率为0.818；基于代谢物数据构建的模型AUC为0.875，准确率为0.833，二者在区分难治性癫痫与非癫痫对照时均表现出较高的分类效能。值得注意的是，整合微生物组与代谢组的分类模型在难治性癫痫与非癫痫对照组中的诊断性能更优，AUC为0.953，准确率为0.875，显著优于单一模型，凸显了多组学整合策略在癫痫诊断中的应用潜力。
结论
本研究结果证实，癫痫患者同时存在肠道菌群失调与脑脊液代谢紊乱，凸显了二者在癫痫发生过程中的协同作用，加深了我们对微生物组-代谢组串扰机制的理解。整合多组学模型展现出更优的诊断性能，强调了其在癫痫精准生物标志物筛选以及癫痫分层与干预的临床应用中的潜力。