基于高通量RNA-Seq和多目标协同演化模因计算的疾病模块识别研究

High-throughput RNA-Seq is becoming the major tool to study gene expression. Reconstruction of genomic co-expression network is pivotal to disease modules identification as well as revealing the underlying interaction mechanism of moleculars. However, due to the large data volume and high resolution nature of RNA-Seq, new efficient gene co-expression network reconstruction and disease modules identification methods oriented to RNA-Seq data are required. In this project, a novel gene co-expression network reconstruction method based on kernel canonical correlation analysis and global indirect correlation silencing technique is proposed toward RNA-Seq data of important diseases. The successive disease modules identification proceeds based on a multiobjective cooperative coevolutionary memetic computing framework. Particularly, the research effort is focused on addressing the compressive storage and access of RNA-Seq data, the evaluation of exon level gene expression correlation, the reconstruction of weighted genomic co-expression network, the definition of multiobjective fitness function, the grouping strategy in large-scale cooperative coevolution, the design of multilevel local search, the ensemble clustering, and the disease modules validation. The outcomes of the project are expected to improve the prediction of disease genes and the understanding of the underlying mechanism of diseases, and provide insight into the system analysis of RNA-Seq data.

高通量RNA-Seq正在成为基因表达研究的主要工具,基于RNA-Seq数据构建基因组共表达网络对于发现疾病模块揭示疾病内部复杂分子机制有重要意义，但由于RNA-Seq的大数据量和高分辨率特性，专门针对其设计的高效共表达网络构建和疾病模块识别方法还非常匮乏。本课题针对重要疾病的RNA-Seq数据，提出核典型关联分析和全局间接相关静默技术相结合的基因共表达网络构建方法，并针对构建网络提出基于多目标协同演化模因计算的疾病模块识别方法。重点解决RNA-Seq数据压缩存取、基于外显子表达水平的基因共表达关系衡量、带权基因组共表达网络构建、多目标网络模块度定义、大规模协同演化分组、多层次局部搜索、集成聚类和疾病模块有效性评估等具体问题，实现针对RNA-Seq数据的高效通用疾病模块识别系统。项目开展有助于提高疾病基因预测水平及对复杂疾病内部分子机制的认识，可为RNA-Seq数据的系统分析提供借鉴。

基于高通量转录组测序（RNA-seq）数据构建基因组共表达网络对于发现疾病模块揭示疾病内部复杂分子机制有重要意义，高通量测序数据的大数据和高分辨率特性，需要高效的存取和分析方法。本课题主要研究内容包括高通量测序数据的高效存取模型研究，基于转录组测序和其它相关基因组测序数据构建疾病复杂网络，并针对构建网络提出基于多目标协同演化模因计算的疾病模块识别方法。项目开展取得了一系列成果包括：1）提出基于参考基因组的高通量测序数据压缩方法，缓解大数据量带来的存储和传输问题；2）针对多种高通量测序数据和疾病相关数据提出新型网络构建技术，发掘疾病相关的网络结构和信息；3）针对全基因组复杂网络模块识别提出分解多目标的协同演化模因计算框架。项目按照计划进度正常执行，执行期间项目组成员在IEEE Transactions on Evolutionary Computation, Bioinformatics, PLoS Computational Biology等重要期刊和国际会议共发表论文29 篇，其中SCI 论文20 篇，中科院一区论文12篇，EI 论文9 篇，一作或者通讯作者23篇，申请发明专利8项，软件著作权1 项，培养博士后1名，硕士研究生7名，超过预期目标。项目取得的研究成果有助于提高疾病基因预测水平及对复杂疾病内部分子机制的认识，可为高通量测序数据的存储和系统分析提供借鉴。