整合先验信息与组学数据重构转录和转录后调控网络的生物信息学方法研究

Reconstructing gene regulatory network from gene expression data is a hot research topic in bioinformatics. While the existing methods should be further improved and enhanced due to the complexity of gene regulation. Transcription factor (TF) and miRNA (microRNA) are two major regulatory factors. Few methods are available so far for reconstructing regulatory relationships containing their combinatorial and synergistic functions. More and more transcriptional regulatory relationships are revealed and documented in literature and databases, such as human regulatory network released in ENCODE project and prior regulations in KEGG. Moreover, multilevel omics datasets, such as transcriptomics of mRNA and miRNA in array profiling and next-generation sequencing, ChIP-Seq, CLIP-Seq and proteomics, indicate the regulatory relationships among the components of TF, miRNA and gene. In this project, we propose to develop novel computational bioinformatics approaches of modeling the comprehensive transcriptional and posttranscriptional regulations by integrating these prior knowledge and heterogeneous datasets. By considering the combination of TF and miRNA regulations in regulatory system, we will identify the time-varying transcriptional regulatory networks during the development process of liver cancer. We will build mathematical programming models for detecting network-based biomarkers and driver modules corresponding to state transitions in complex disease, such as liver cancer. Thus the project will generate new models and algorithms, software and database for computational biology research community, and also will provide new perspective and insight for describing the pathogenesis of hepatocellular carcinoma and complex regulatory mechanisms.

基因调控网络重构是生物信息计算中的热点问题。转录因子TF和miRNA是两大类重要调控因子，当前甚少有同时考虑TF和miRNA组合协同作用的转录和转录后调控网络模型；文献和数据库中已收录越来越多的调控关系，如ENCODE计划公布的转录调控和KEGG数据库记录的先验调控知识；同时，多层次组学数据如基因/miRNA表达谱、RNA/ChIP/CLIP-Seq、PPI等提供了有关TF、miRNA和基因之间调控关系的组学数据。本项目旨在开发整合这些先验信息与多层次数据的重构转录和转录后调控网络的模型和方法。通过考虑TF和miRNA在调控系统中的组合协同作用，整合先验信息和异源数据，开发综合调控网络重构模型；推断肝癌发生发展中动态时变调控网络，解析基因调控和表观调控的协同；探测疾病状态变化中的调控网络标记物和解析致病机理。从而为生物信息学研究提供新模型和方法、软件和数据库，为描述基因调控现象提供新视角。

本项目主要开展基于多分子层次组学数据的基因调控网络重构及应用方面的研究工作。首先，针对TF-miRNA组合调控的重要性，项目组整理出TF-miRNA协同作用的基因调控网络数据库RegNetwork，并提供网站服务；针对先验调控信息在基因调控网络推断中的重要意义，项目组给出了包含先验知识的必要性，特别是包含miRNA介导的转录后调控网络对于网络推断可靠性与真实性的影响；给出了利用相关性方法构建基因调控网络的系统比较研究，给出了相关网络与因果网络之间的区别与联系。其次，项目组给出非编码RNA与蛋白质相互作用的机器学习预测方法；给出蛋白质表面绑定RNA局部结构间的比对算法并系统研究了蛋白质绑定RNA的主要结构模式，解析了基因调控特别是转录后调控的结构基础。再次，项目组基于上述基因调控网络模型，给出了基于动态熵函数、差异相互作用及改进PageRank的生物标记物发现模型与方法，利用多分子层次组学数据分别得到了肝癌、糖尿病以及阿尔兹海默症等复杂疾病可能的生物标记物。项目组从基因调控网的构建、结构绑定基础以及基于差异网络模型的生物标记物发现三个层次，系统地对整合多分子层次组学数据的基因调控网络生物信息学模型与方法进行了研究，取得一系列进展，共发表论文17篇，其中SCI收录12篇，EI收录4篇，申请国内发明专利1项，获2017年教育部自然科学二等奖1项。本项目为生物信息学研究提供新模型和方法、新软件和数据库，并从计算的角度为研究基因调控现象提供了新工具和新视角。