样本特异性基因功能模块挖掘的多源信息融合方法研究

It is a hot and difficult point in the research of life science to excavate the functional modules related to human complex diseases, especially for the specificity of samples to realize the personalized medical treatment of diseases. Aiming at the problem of digging disease-related functional modules, the project intends to focus on how to fully extract different omics information with sample specificity and that of gene functional similarity, design fusion strategies, and find appropriate formal models for digging disease-related modules. The main contents include: The extraction of the disease-related genes with sample specificity at different levels of omics using deep neural network, information entropy, weighted distance matrix and statistical methods; By using the probabilistic topic model and the graph structure analysis method, the semantic association information between genes is extracted and the gene function similarity network of multi-data integration is constructed; The heterogeneous gene network is constructed based on the similarity network fusion theory through fusing the information of sample specificity and gene functional similarity network, and then the related functional modules with sample specificity are mined. The research of the project is based on the simulation data and the real data of multi-diseases and multi-omics. The research results will help explain the pathogenesis of the disease with sample specificity and provide help for the risk assessment, early diagnosis and personalized treatment of diseases.

挖掘人类复杂疾病相关的功能模块，尤其是针对样本特异性实现疾病的个性化医疗，是生命科学研究的热点和难点。本项目针对疾病相关功能模块的挖掘问题，拟重点研究如何充分提取样本特异性的不同组学信息和基因功能相似性信息，设计融合策略，寻找合适的形式化模型实现疾病相关功能模块的挖掘。主要研究内容包括：综合利用深度神经网络，信息熵及加权距离矩阵和统计方法，有效提取不同组学层面下基因的样本特异性信息；利用概率主题模型，结合图结构分析方法，提取基因间的语义关联信息，构建多数据集成的基因功能相似网络；基于相似网络融合理论将样本特异性信息和基因功能相似网络进行融合，构建异质基因网络，进而挖掘样本特异性疾病相关功能模块。本项目研究基于仿真数据和多疾病多组学真实数据展开，研究成果将有助于解释样本特异性的疾病致病机理，为疾病的风险评估、早期诊断和个性化治疗提供帮助。

复杂疾病（如癌症）严重危害人类身体健康，其发生和发展往往是多分子相互作用导致的功能失调所引起，因此，挖掘人类复杂疾病相关的功能模块，尤其是针对样本特异性实现疾病的个性化医疗，是生命科学研究的热点和难点。本项目研究样本特异性的复杂疾病致病机理，整合生物多组学、基因本体（Gene Ontology, GO）等多源数据的样本特异性信息提取、基因相似性网络构建以及样本特异性的基因及功能模块识别等。在样本特异性信息提取方面，提出了一种基于信息增益的特异性信息提取算法，并进一步利用特异性信息对样本进行分层，研究了癌症的异质性机理；在基因相似性网络构建方面，通过融合GO注释信息，提出了基于网络表示学习方法的基因网络相似性预测方法，并利用链接预测验证所构建网络的可靠性；在复杂疾病关联基因及功能模块识别方面，提出了基于基因贡献得分的疾病亚型识别方法，并进一步识别亚型相关基因和模块，并利用生存分析和富集分析验证所识别基因和功能模块的预后价值。本项目的完成为探索样本特异性的复杂疾病相关基因和功能模块挖掘提供了新的思路和方法，并进一步解释样本特异性的疾病致病机理，为疾病的风险评估、早期诊断和个性化治疗提供帮助。