基于新一代测序数据的复杂疾病特异共调控网络构建及分析方法研究

The pathogenesis of complex diseases is a complicated process which involves multiple factors such as sequences, genes, non-coding RNAs, proteins and epigenetics. How to interpret the correlations among these factors and their combinatorial effects towards complex disease is one of the most difficult challenges at present. The emergence and rapid growth of the high-throughput next-generation sequencing techniques has provided new opportunities for analyzing the pathogenesis of complex diseases with multi-omics data. On the basis of the next-generation sequencing data, transcriptomics data and protein-protein interaction data, and with the help of computational methods such as data mining, machining learning as well as biostatistics, we will design new methods to identify differentially expressed factors with sparse representation and construct disease-specific co-regulatory networks with multi-omics data. Furthermore, we will use graph theory-based methods to analyze the topological and functional characteristics of the disease-specific co-regulatory networks, discover disease-related topological substructures, study the specific or relevant disease expressions, eventually reveal the regulatory elements, regulatory patterns and the regulatory pathways that are closely related to diseases. The accomplishment of this project will provide new insights into the multi-omics data integration at the network level and the application of complex disease analysis; it will also provide the scientific basis of finding potential drug targets for complex diseases as well as accelerating the pace of drug development process.

复杂疾病的发病过程涉及序列、基因、非编码RNA、蛋白质及表观遗传等多因素作用，如何阐明多因素间的关联作用及其在复杂疾病发生发展过程的影响是目前针对复杂疾病研究的一大难题。新一代高通量测序技术的出现为多组学融合的复杂疾病致病机理研究提供新的机遇。本项目以新一代高通量测序数据、转录调控数据及蛋白质相互作用数据等为基础，以数据挖掘、机器学习和生物统计等计算方法为手段，设计基于稀疏表示的差异表达因子识别方法及疾病特异的多组学调控网络构建方法，采用基于图论的方法分析疾病特异的共调控网络的拓扑特性和功能特性，发现与复杂疾病相关的网络拓扑子结构，研究复杂疾病特异性或相关性表达，进而揭示与疾病发生密切相关的调控元件、调控模式及调控通路。本项目的完成将为基于网络水平的多组学数据融合及其在复杂疾病分析中的应用研究提供新的思路，同时为寻找治疗复杂疾病的潜在药物靶点，加快药物研究进程提供网络水平上的科学依据。

鉴于复杂疾病的发生发展涉及到基因、非编码RNA、蛋白质及表观遗传等多因素的作用，本项目基于复杂疾病相关的基因组、转录组、蛋白质组等组学数据，通过形成一套完整的共调控网络分析体系，在网络水平上对复杂疾病的发病机制、演化模式及作用规律展开了研究。通过利用新一代疾病特异的数字化表达谱数据识别差异表达的致病因子miRNA/mRNA，结合其他相关组学数据信息构建了有效可靠的疾病特异共调控网络，并对疾病特异共调控网络中关键节点、网络模体及功能模块的识别问题进行了研究，以深度挖掘共调控网络中蕴含的生物医学信息。针对现有关键节点识别算法准确度不高且难以直接应用于疾病特异共调控网络的问题，结合共调控网络特有拓扑结构及其他生物特征信息，提出了基于线性模型的关键共调控节点识别算法co-BOTLM、基于差异表达基因和关键因子调控能力的关键因子并行识别模型MRDGC，有效提高了识别精度；为克服共调控网络中识别模体计算复杂度高的问题，提出基于G-tries结构的大型共调控模体识别算法LCNM，降低了计算开销；针对共调控网络的稀疏性及多类型节点的差异性问题，通过整合共调控网络拓扑特性及多组学数据信息，提出了多种诸如SNCoNMF、OSC、APNICRM的共调控模块识别算法，不仅提高了识别精度，还能够挖掘出更多具有生物意义的共调控功能模块。再者，为更深层次地理解复杂疾病的发病机制，研究了多种复杂疾病与miRNA、mRNA、circRNA及微生物等各类分子之间的作用机制，通过结合机器学习与网络学习方法，提出了致病miRNA预测模型GRNMF、致病靶基因预测算法IMTRBM及致病circRNA预测算法MRLDC等一系列致病因子预测及关联分析模型，且均取得良好的预测性能，为更好地在疾病特异共调控网络中准确定位关键致病因子、调控模式及调控通路提供了有力支撑。