Abstract:From transcriptional noise to dark matter of biology, the rapidly changing view of long non-coding RNA (lncRNA) leads to deep understanding of human complex diseases induced by abnormal expression of lncRNAs. There is urgent need to discern potential functional roles of lncRNAs for further study of pathology, diagnosis, therapy, prognosis, prevention of human complex disease and disease biomarker detection at lncRNA level. Computational models are anticipated to be an effective way to combine current related databases for predicting most potential lncRNA functions and calculating lncRNA functional similarity on the large scale. In this review, we firstly illustrated the biological function of lncRNAs from five biological processes and briefly depicted the relationship between mutations or dysfunctions of lncRNAs and human complex diseases involving cancers, nervous system disorders and others. Then, 17 publicly available lncRNA function-related databases containing four types of functional information content were introduced. Based on these databases, dozens of developed computational models are emerging to help characterize the functional roles of lncRNAs. We therefore systematically described and classified both 16 lncRNA function prediction models and 9 lncRNA functional similarity calculation models into 8 types for highlighting their core algorithm and process. Finally, we concluded with discussions about the advantages and limitations of these computational models and future directions of lncRNA function prediction and functional similarity calculation. We believe that constructing systematic functional annotation systems is essential to strengthen the prediction accuracy of computational models, which will accelerate the identification process of novel lncRNA functions in the future.

Computational models for lncRNA function prediction and functional similarity calculation.