基于高通量数据的基因调控网络构建模型和方法研究

An important problem in bioinformatics and systems biology is to identify and understand the gene regulatory networks (GRNs), which explicitly represent the regulatory process in cell systems. Reconstruction of GRN remains a bottleneck in the pipeline to discover the regulatory mechanism from high-throughput data. The project will apply and develop novel computational models and algorithms to construct transcription factor, microRNA and target gene included hierarchical network from high-throughput data such as next-generation sequencing and microarray data. There are several key works will be finished in the project. Firstly, a two-level optimization model to filter out the indirect regulations will be proposed and a novel GRN inference algorithm based on the proposed model will be developed. Secondly, a prior and posterior information-constrained recursive optimization model will be proposed to reduce the noisy and redundant links in a GRN. Based on the proposed model, a computational algorithm will be developed to improve the accuracy of GRN inference. Thirdly, quantitative measures for distinguishing different combinatorial regulations (i.e. synergy, competitive and indirect regulations) will be proposed based on information theory. A computational algorithm and software will be developed to construct transcription factor and microRNA-mediated transcription and post-transcription regulatory network from sample matched mRNA and microRNA RNA-Seq data.

生物信息学与系统生物学中一个重要任务是认识和理解能够代表某种生物调控机制的基因调控网络。基因调控网络构建是通过高通量生物数据挖掘生物调控机理的是一个重要瓶颈。本项目将集成如新一代测序、基因芯片等大规模高通量数据，利用计算模型构建转录因子、microRNA和靶基因等多层次上的调控网络，开发高效实用的基因调控网络构建算法和软件。主要包括，(1)提出能够滤除间接调控关系的双层优化模型，并基于此模型来开发高准确率的基因调控网络构建算法和软件。(2)集成调控关系数据信息，提出基于先验信息和后验信息双重约束的逐步优化模型，开发基因调控网络构建的高精度降噪去冗余算法。(3)基于信息理论提出区分组合调控关系(如协同调控、竞争调控和间接调控)的定量模型和算法，并集成同样本的mRNA和microRNA的RNA-Seq数据，开发转录因子和microRNA介导的转录和转录后调控网络的构建算法和软件。

在过去三年，本课题针对“基因调控网络构建”开展了深入的研究，提出了多个新颖、高效的算法和模型，整个课题按照研究计划执行。在该课题资助下，课题组在以下几个方面取得了突破性进展：（1）基于信息理论提出了一种准确定量调控关系和强度的条件互容信息模型CMI2，建立了基于小样本推断基因调控网络的新模型和算法CMI2NI；（2）基于PCA-CMI方法的顺序依赖性缺点，即网络节点的不同排列顺序可能得到不同的网络结构，开发了基因调控网路构建的鲁棒性算法CNMIT；（3）在模型CMI2的基础上开发了能够定量直接关联的偏互信息模型PMI，并基于此模型的基因调控网络构建算法PCA-PMI；（4）通过并行计算框架对测量基因调控关系的条件互信息进行计算，开发了能够构建全基因组调控网络的软件包CMIP;（5）提出基于先验信息和后验信息双重约束的逐步优化模型，开发基因调控网络构建的高精度降噪去冗余算法。在该项目资助下，课题组发表论文6篇，其中SCI收录期刊4篇。发表的杂志包括生物信息学顶级杂志Proceedings of the National Academy of Sciences，USA、Nucleic Acids Research、BMC Bioinformatics、Molecular BioSystems等。其中发表的论文还被世界学术组织F1000Prime列为推荐论文。