Identification of drug resistance associated ncRNAs based on comprehensive heterogeneous network

Aims: Chemotherapy and molecularly targeted therapy are main strategies for treatment of cancers. However, long-term treatment makes cancer cells acquire resistance to anti-cancer drugs, which severely limits the effects of cancer treatment. NcRNAs, especially miRNAs and lncRNAs, have been reported to play key roles in drug resistance and could restore drug responses in resistant cells. Main methods: We presented a network-based framework to systematically identify drug resistance associated miRNAs and lncRNAs. First, we constructed a comprehensive heterogeneous miRNA-lncRNA regulatory network through integrating curated miRNA regulations to lncRNA, and significantly co-expressed miRNA-miRNA, lncRNA-lncRNA and miRNA-lncRNA interactions for each cancer type. Second, random walk with restart (RWR) was utilized to identify novel drug resistance associated ncRNAs. Key findings: We predicted 470 associations of 34 miRNAs and 79 lncRNAs for 27 drugs in 10 cancer types. In addition, leave-one-out cross validation (LOOCV) demonstrated the effectiveness of the proposed approach. Next, we also demonstrated that the integrated heterogeneous cancer-specific network achieved better performance than the general curated miRNA-lncRNA regulatory network. What's more, we found that the drug resistance associated ncRNAs validated by high-throughput technology was also a reliable source for prediction. Significance: We proposed a new framework to identify novel and reliable drug resistance associated ncRNAs, which provides new perspectives for drug resistance mechanism and new guidance for clinical cancer treatment.