Sequence-based prediction of microRNA transcription start sites from transcriptome data using DNA structural properties

• Main Authors: Yu-Ming Hsiao, 蕭鈺銘
• Other Authors: 謝立青
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/57u4de

碩士 === 國立中興大學 === 基因體暨生物資訊學研究所 === 106 === microRNAs (miRNAs) are important small non-coding RNAs, mainly regulating gene expression in cells. Understanding upstream promoter regions of miRNAs is also an important issue. Current developed methods to identify miRNA transcription start sites (TSSs) all require specifically designed experiments or integrated multiple data. Therefore, we want to construct a more generic method. We use standard RNA sequencing (RNA-seq) data and sequence features (including conservation score, CpG content, and structural properties), and utilize the Support Vector Machine (SVM) maching learning algorithm to predict miRNA TSSs. Based on the method we have established before, this time we incorporate the structural properties into models, which improve the performance of the models. Comparison of our prediction results with experimentally validated TSS data shows that our method can identify miRNA TSSs. Moreover, this prediction system is not limited to human or mouse. It can be applied to more species because of its generic characteristics in sequence features. We hope these predicted TSSs can compensate for the lack of experimentally validated TSS data or provide TSS candidate sites for further experimental verification, to comprehensively understand the role of miRNAs in regulatory network.