Model Construction For MicroRNA Biomarker Discovery

MicroRNAs(miRNAs) are a class of small non-coding RNAs with approximately 22-24 nucleotides in length. They regulate gene expression at the post-transcriptional level. Plenty of researches indicated that miRNAs were functional in many important biological processes and could be effective biomarkers for disease diagnosis and treatment.Currently, a lot of approaches have been reported on detecting miRNAs as biomarkers. Most of them are experimental, which may be time-consuming and costly. Bioinformatics models are therefore preferred. Unfortunately, they are not generalized enough. Until now, no general rules can be utilized for miRNA biomarker discovery.In this study, we reconstructed human miRNA-mRNA network, paying more attention to its fragile sites and making in-depth analysis on the independently regulatory power of miRNAs. Combining with biological functions and gene evolutionary patterns, statistical results showed that biomarker miRNAs were likely to regulate more transcription factor genes. Thus, we constructed the computational model for mi RNA biomarker discovery based on statistical evidences, and implemented it in the program Micro RNA Biomarker Discovery(termed Mi RNA-BD). Compared with traditional methods which highly rely on the training data, our model detects miRNA biomarkers precisely without any prior knowledge. On the other hand, applications to pediatric acute myeloid leukemia as well as other complex diseases demonstrated its great predictive power.The study provides theoretical rules and techniques for mi RNA biomarker discovery, which is significant and valuable for scientific researches.