Characterization of long non-coding RNA-mediated competing endogenous RNA network to reveal potential long non-coding RNA biomarkers in rheumatoid arthritis patients under anti-TNF treatment

Objective: We designed to identify long non-coding RNA (lncRNA) signatures that predicted the therapy outcomes in rheumatoid arthritis (RA) via characterizing lncRNA-mediated competing endogenous RNA network (LMCN). Methods: Hypergeometric test was used to detect the competing lncRNA-mRNA interactions, following by co-expression analysis for the competing lncRNA-mRNA interactions relying on pearson correlation coefficient (PCC). The PCC absolute value of one interaction was defined as the weight value of one edge. Using the weight value threshold of 0.6, we established a highly competitive LMCN. To analyze the network organization, we conducted degree distribution of LMCN. Moreover, to predict the function of significant lncRNAs, we implemented the pathway analysis for its mRNA neighbors in the LMCN. Finally, we employed Biclique algorithm to extract competing modules from the LMCN. Results: Relying on a weight value > 0.6, a high-competing LMCN was constructed, which covered 50 lncRNAs, 824 mRNAs, and 926 competing endogenous (ceRNA) interactions. Based on degrees > 40, a total of 7 hub genes were identified, including SNHG12, C1RL-AS1, TTTY15, MALAT1, TAPT1-AS1, LINC00476, and LINC00649. Remarkably, a hub lncRNA LINC00476 was involved in the signaling-associated pathways (phosphatidylinositol signaling system, Ras signaling pathway, Rap1 signaling pathway, and TNF signaling pathway). Overall, we extracted 1 synergistic, competitive module containing 30 nodes. Significantly, one lncRNA LINC00476 in the LMCN was also the hub within the module. Conclusion: Our exploratory study indicates that the obtained lncRNA signatures provide novel information to better understand the mechanisms of action of anti-TNF treatment in RA patients.