Data collection and preprocessing
To determine whether there was clustering or outliers in the sample set, the differences between the clustering of the mRNA (Fig. 1A–1C), lncRNA (Fig. 1D–1F), and miRNA (Fig. 1G, 1H) expression matrixes of the NPC and normal tissue samples in different datasets were examined using three‐dimensional principal component analysis (PCA). The results showed that NPC was well distinguished from the normal tissue samples.
Identification of DEGs in NPC
To identify the DEGs in NPC, the mRNA, lncRNA, and miRNA expression profiles were analyzed using the Limma package. The results showed that 3664 mRNAs, 4068 lncRNAs, and 265 miRNAs were dissimilarly expressed (|logFC| ≥ 1, adjusted P ≤ 0.01; Fig. 2) between the NPC and normal tissues. Of these, 2181 mRNAs, 2087 lncRNAs, and 175 miRNAs were significantly upregulated, while 1483 mRNAs, 1981 lncRNAs, and 90 miRNAs were significantly downregulated. In total, 61 DEGs were shared among the three mRNA datasets (Fig. 2D), 10 differentially expressed lncRNAs (DElncRNAs) were shared among the three lncRNA datasets (Fig. 2H), and 14 differentially expressed miRNAs (DEmiRNAs) were shared among the two miRNA datasets (Fig. 2K). Those DEGs may provide new insight into the biological mechanisms of NPC and serve as potential therapeutic targets for NPC. Functional roles of 10 DElncRNAs shared among the three lncRNA datasets are provided in Table S1, and functional roles of 14 DEmiRNAs shared among the two miRNA datasets are provided in Table S2.
Construction of the ceRNA network
To explore the role of miRNAs and corresponding target mRNAs, as well as corresponding lncRNAs in NPC, we predicted the target mRNAs of the DEmiRNAs, and lncRNAs that may have interrelations with miRNAs. The results may help to better explain the critical regulatory functions of miRNAs, mRNAs, and lncRNAs. The interaction of up-regulated and down-regulated miRNAs with DElncRNAs was predicted based on miRcode. The prediction of target mRNAs of up-regulated and down-regulated miRNAs was performed using three databases, miRDB, miRTarBas, and TargetScan. LncRNAs, miRNAs and mRNAs were included in the up-regulated and down-regulated lncRNA-miRNA-mRNA ceRNA networks respectively (Fig. 3A, 3B). The blue, red and green nodes represent miRNAs, lncRNAs, and miRNAs, respectively. Table S3 and Table S4 shows the details of the interactions of the up-regulated and down-regulated miRNAs and mRNAs, respectively. Table S5 and Table S6 shows the details of the interactions of the up-regulated and down-regulated miRNAs and lncRNAs, respectively.
GO and KEGG analyses of DEGs
To further analyze the possible functions of the 61 DEGs, we conducted biological analyses by using clusterProfiler and UpSetR. The results suggested that the DEGs were significantly enriched in GO and KEGG terms. The GO analysis showed that the following BPs were notably enriched among the DEGs: chromosome segregation, nuclear chromosome segregation, sister chromatid segregation, mitotic sister chromatid segregation, negative regulation of chromosome organization (Fig. 4A). The following MFs were largely enriched in ATPase activity, protein serine/threonine kinase activity, ATPase activity (coupled), tubulin binding, catalytic activity (acting on DNA), DNA dependent ATPase activity, DNA helicase activity, and single-stranded DNA dependent ATPase activity (Fig. 4B). Finally, the following CCs were found to be largely enriched in the chromosomal region, condensed chromosome, chromosome (centromeric region), condensed chromosome (centromeric region), nuclear chromosome (telomeric region), and condensed chromosome kinetochore (Fig. 4C). The KEGG pathway analysis suggested that DEGs in NPC were largely enriched in the cell cycle, DNA replication, and small cell lung cancer (Fig. 4D). The results suggested that chromosomal dysfunction was closely related to the development of NPC.
GSEA of NPC-related genes
To explore the biological functions of the DEGs involved in NPC, GSEA was applied. The mRNA expression profile of the FJCH dataset was subjected to GSEA by means of clusterProfiler. The analysis showed that the following biological pathways were over-represented in the NPC tissues as compared to the normal tissues : the mitogen-activated protein kinase (MAPK) signaling pathway, the phosphatidylinositol-3 OH kinase/protein kinase B (PI3K-Akt) signaling pathway (Fig. 5A), the apoptotic pathway, and the tumor necrosis factor (TNF) signaling pathway (Fig. 5B). The pathways found our study were involved with cancer progression, metastasis, and apoptosis.
PPI network analysis of DEGs
The STRING database was used (version: 11.3) to explore the PPI network based on the correlations among the 61 DEGs in NPC. The obtained data were then examined using Cytoscape software. The PPI network of DEGs was constructed using MCODE to obtain the vital gene module. The NetworkAnalyzer plugin was applied to further analyze the PPI network according to the scores. The cytoHubba plugin was used to analyze the hub genes associated with NPC, and the following genes with the top 20 grades were deemed to be hub genes: NUSAP1, RACGAP1, PRC1, KIF4A, TOP2A, PBK, KIF2C, TPX2, CENPU, OIP5, TTK, MAD2L1, NDC80, BIRC5, MELK, CENPF, FOXM1, TYMS, CDK1, and CEP55 (Fig. 6). Those genes may contribute to the investigation of biological mechanisms and uncover underlying therapeutic targets for NPC.
GO and KEGG analyses of hub genes
To analyze the functions of the top 20 hub genes, we again conducted biological analyses by using clusterProfiler and UpSetR. The results suggested that the hub genes were significantly enriched in GO and KEGG terms. GO analysis showed that changes in the following BPs of hub genes were notably enriched in chromosome segregation, nuclear chromosome segregation, sister chromatid segregation, mitotic sister chromatid segregation, microtubule cytoskeleton organization involved in mitosis, and regulation of chromosome segregation (Fig. 7A). In addition, the changes in the following MFs were mainly enriched in protein serine/threonine kinase activity, tubulin binding, microtubule binding, and protein C-terminus binding (Fig. 7B). Finally, changes in the following CCs of DEGs were enriched in the chromosomal region, spindle, condensed chromosome, chromosome (centromeric region), kinetochore, microtubule, midbody, condensed chromosome (centromeric region), condensed chromosome kinetochore, and mitotic spindle (Fig. 7C). KEGG pathway analysis indicated that the DEGs in NPC were mainly enriched in the cell cycle, cellular senescence, oocyte meiosis, progesterone-mediated oocyte maturation, and platinum drug resistance (Fig. 7D). Enrichment analyses of the hub genes were similar to the results of the analyses of the DEGs. Hence, the findings obviously suggested that chromosomal dysfunction was a vital contributor to the tumorigenesis of NPC.