Research On Disease-related MicroRNA Based On Biological Heterogeneous Networks

With the complement of the Human Genomes Sequencing,there is a dramatic increase in the number of biological data.Conventional bio-test methods cannot keep pace with the recent research needs and then bioinformatics is a new gradually rising interdisciplinary that develops computational methods and software tools to detect biological knowledge.Therefore,there is an inevitable trend using computational methods to explore potential interactions and to solve biological problems in future research.This thesis focuses on link prediction between miRNAs and diseases.MiRNAs are a kind of important small RNA molecules,which play crucial regulatory roles in a cell and imperative roles in the occurrence and development of diseases,especially cancers.Predicting potential disease-related miRNAs can help to understand the pathogenesis of diseases from the post-transcriptional levels,thus to support clinical and medical research.This thesis investigates prediction methods based on biological heterogeneous networks and mainly includes three following aspects:(1)Integrates kinds of latest data,uses some methods to computer the similarity,computes miRNAs and diseases similarity networks,and finally reconstructs the miRNA-disease heterogeneous networks.(2)Utilizes the method based on meta-path to predict potential associations between miRNAs and diseases and compares the methods with five other classical prediction methods to evaluate the performance of prediction.(3)Proposes two improvements based on meta-path algorithm.One is considering the miRNAs' labels by setting different threshold for miRNAs to improve precision of prediction.Other is using Support Vector Machine to learn different coefficient of each meta-path to improve methods.The performance of our improved methods is superior to the existing methods.