We aimed to investigate the associations between the intestinal microbiota, metabolites, cytokines, and clinical severity in anti-N-methyl-d-aspartate receptor (NMDAR) encephalitis and to further determine the predictive value of the intestinal microbiota or metabolites in clinical prognosis.In this prospective observational cohort study of 58 NMDAR encephalitis patients and 49 healthy controls, fecal microbiota, metabolites, and cytokines were quantified and characterized by16S rRNA gene sequencing, liquid chromatography-mass spectrometry, and the Luminex assay, respectively.There were marked variations in the gut microbiota composition and metabolites in critically ill patients. We identified 8 metabolite modules (mainly characterized by fatty acid, glycerophosphoethanolamines, and glycerophosphocholines) that were distinctly classified as negatively or positively associated with bacterial co-abundance groups (CAGs). These CAGs were mainly composed of Bacteroides, Eubacterium_hallii_group, Anaerostipes, Ruminococcus, Butyricicoccus, and Faecalibacterium, which were substantially altered in patients. In addition, these fecal and serum metabolic modules were further correlated with the serum cytokines. Additionally, the combination of clinical features, microbial marker (Granulicatella), and a panel of metabolic markers could further enhance the performance of prognosis discrimination significantly, which yielded an area under the receiver operating characteristic curve of (AUC) of 0.94 (95%CI = 0.7-0.9). Patients with low bacterial diversity are more likely to develop relapse than those with higher bacterial diversity (log-rank p = 0.04, HR = 2.7, 95%CI = 1.0-7.0).The associations between the multi-omics data suggested that certain bacteria might affect the pathogenesis of NMDAR encephalitis by modulating the metabolic pathways of the host and affecting the production of pro-inflammatory cytokines. Furthermore, the disturbance of fecal bacteria may predict the long-term outcome and relapse in NMDAR encephalitis.Copyright © 2022 Gong, Liu, Liu, Li, Guo, Zhou and Hong.