Mining The Functional RNA Methylation Genes

As the most prevalent mammalian m RNA epigenetic modification,N~6-methyladenosine(m6A)is involved in regulating many different aspects of m RNA metabolism and diseases like cancer.Me RIP-Seq high throughput sequencing data analysis can help reveal the potential roles of m RNA m6A methylation in regulation of gene expression,splicing and so on.However,the existing Me RIP-Seq data analysis ignore the functional information of m6A.Thus,it is eager to develop effective computational methods and models for detecting functional m6A methylation genes to help finding the m6A function and to study the regulation mechanism of m6A in important biological pathways and diseases.In this thesis,the methods of m6A methylation and differential m6A methylation site prediction,m6A-driven gene and functional differential m6A methylation(Dm M)gene identification and m6A-associated diseases analysis are studied.The main works are as follows:1.The existing m6A peak detection methods from Me RIP-Seq data can only identify condition specific m6A site with limited low resolution and high false positive rate.Considering this,we integrated the single-base m6A site information from mi CLIP data and proposed a novel convolutional neural network(CNN)model,Deep-m6A,to predict condition specific m6A site in single-base resolution by taking the Me RIP-Seq data and RNA sequence information as input.Furthermore,a single-base Dm M site prediction algorithm,DMDeep-m6A,was proposed based on Deep-m6A and rhtest.The results show that Deep-m6A outperforms the existing algorithms,which indicates that Deep-m6A can effectively identify condition specific m6A methylation sites with reduced false positive rate.2.The existing Me RIP-Seq data analysis algorithms can only identify Dm M sites and genes in isolation,which neglect the functional interaction between Dm M genes.To sovle this,we proposesd m6A-Driver to identify m6A-driven genes whose functional interactions are likely to be actively modulated by m6A methylation under a specific condition based on random walk with restart(RWR)algorithm.The results on 4 known m6A(de)methylases,i.e.,FTO,METTL3,METTL14 and WTAP datasets suggest that m6A-Driver can robustly and efficiently identify m6A-driven genes that are functionally more enriched and associated with higher degree of differential expression than Dm M genes.Pathway analysis of the constructed context-specific m6A-driven gene networks further revealed the regulatory circuitry underlying the dynamic interplays between the methyltransferases and demethylase at the epitranscriptomic layer of gene regulation.3.To reveal the dynamic changes of m6A levels in specific context and the role of the changes in certain biological processes,we proposed Fun DMDeep-m6A in this work to identify and prioritize functional Dm M genes(FDm MGenes).This proposed network method includes a novel m6A-signaling bridge(MSB)score to quantify the functional significance of Dm MGenes by assessing functional interaction of Dm MGenes with their signaling pathways using a heat diffusion process in protein-protein interaction(PPI)networks.The test results on 4 context-specific Me RIP-Seq datasets showed that Fun DMDeep-m6A can identify more context-specific and functionally significant FDm MGenes than m6A-Driver.The functional enrichment analysis of these genes revealed that m6A targets key genes of many important context-related biological processes.These results demonstrate the power of Fun DMDeep-m6A for elucidating m6A regulatory functions and its roles in biological processes and diseases.4.Our current knowledge about how m6A levels are controlled,whether and how regulation of m6A levels of a specific gene can play a role in cancer and other diseases is mostly elusive.To study this,we proposed in this paper a computational scheme for predicting m6A-regulated genes and m6A-associated disease,which includes Hot-m6A,a new network-based pipeline that prioritizes functional significant m6A genes and the application of RWRH to identify m6A-associated diseases using gene-disease heterogeneous networks.The functional enrichment analysis of these genes and networks demonstrate the power of our proposed computational scheme and provide new leads for understanding m6A regulatory functions and its roles in diseases.