Background: Similar to that in other malignant tumors, distant metastasis is one of the most important causes of poor prognosis in nasopharyngeal carcinoma (NPC). However, the genetic hallmarks and networks that regulate the distant metastasis of NPC are not fully understood. Methods: In this study, we performed high-throughput screening of mRNA expression profiles in 92 NPC samples collected from 3hospitals and detected the mRNA expression levels of 31,503 genes in these samples. Gene functional enrichment analyses were performed using gene set enrichment analysis (GSEA). Least absolute shrinkage and selection operator (LASSO) was applied to select prognostic genes and a Cox proportional hazards regression model including these genes was constructed to predict prognosis. The Kaplan-Meier curve and time-dependent receiver operating characteristic (ROC) curve were plotted to assess the performance of this model. Univariate and multivariate analyses were performed using the Cox proportion hazard model to test the independence of prognostic effect of gene model and other clinical features. Results: A total of 1837 differentially expressed genes between patients with and without distant metastasis were identified in the training cohort, including 869 upregulated genes and 968 downregulated genes. Six gene sets, including the Wnt/beta catenin signaling pathway, hedgehog (Hh) signaling pathway, Notch signaling pathway, mitotic spindle, apical surface, and estrogen response late, were enriched in patients with distant metastasis. A four-gene signature model was constructed in the training cohort, and according to the time-dependent ROC curve, this model had certain accuracy in predicting distant metastasis-free survival (DMFS) in both the training and validation cohorts. Conclusion: We developed a four-gene signature model that can evaluate the distant metastasis risk of NPC patients and may also provide novel therapeutic targets for NPC treatment in the near future.