Integrative Clustering Reveals a Novel Subtype of Soft Tissue Sarcoma With Poor Prognosis

BACKGROUND: Soft tissue sarcomas (STSs) are heterogeneous at the clinical and molecular level and need to be further sub-clustered for treatment and prognosis. MATERIALS AND METHODS: STSs were sub-clustered based on RNAseq and miRNAseq data extracted from The Cancer Genome Atlas (TCGA) through the combined process of similarity network fusion (SNF) and consensus clustering (CC). The expression and clinical characteristics of each sub-cluster were analyzed. The genes differentially expressed (lncRNAs, miRNAs, and mRNAs) between the poor prognosis and good prognosis clusters were used to construct a competing endogenous RNA (ceRNA) network. Functional enrichment analysis was conducted and a hub network was extracted from the constructed ceRNA network. RESULTS: A total of 247 STSs were classified into three optimal sub-clusters, and patients in cluster 2 (C2) had a significantly lower rate of survival. A ceRNA network with 91 nodes and 167 edges was constructed according to the hypothesis of ceRNA. Functional enrichment analysis revealed that the network was mainly associated with organism development functions. Moreover, LncRNA (KCNQ1OT1)-miRNA (has-miR-29c-3p)–mRNA (JARID2, CDK8, DNMT3A, TET1)-competing endogenous gene pairs were identified as hub networks of the ceRNA network, in which each component showed survival significance. CONCLUSION: Integrative clustering analysis revealed that the STSs could be clustered into three sub-clusters. The ceRNA network, especially the subnetwork LncRNA (KCNQ1OT1)-miRNA (has-miR-29c-3p)-mRNA (JARID2, CDK8, DNMT3A, TET1) was a promising therapeutic target for the STS sub-cluster associated with a poor prognosis.