基于异质多组学数据集成的基因调控网络建模方法研究

Genes together with their products, which are RNAs or proteins generated from expression of genes, are interacting with each other and thereby forming gene regulatory networks. Various biological functions can be enabled by these gene regulatory networks, and they play important roles in human lives. Recently rapid development of high-throughput technologies brought large-scale multi-omics data. How to efficiently extract useful information from these data for constructing and analyzing gene regulatory network is meaningful for uncovering relevant functional mechanism. In this project, we will focus on modeling gene regulatory networks via integrating multi-omics data. Firstly, based on the technique of network decomposing and combining by turns, we will propose a method named local Bayesian networks to reconstruct the directed regulatory networks in multi-omics data. Secondly, to reduce the impact of insufficiency of labeled data on supervised learning method, we shall propose a reinforcement learning based method to reconstruct the undirected regulatory networks in multi-omics data. Thirdly, we will study the regulatory networks of epigenetic modification by network propagation with resistance. The outcomes of this project could be helpful for biologist and medical scientists to intuitively analyze the complex regulatory mechanisms underlying gene regulatory networks and provide technical support for further personalized medicine and the design of novel drugs.

基因及其表达过程中形成的RNA或蛋白质等基因产物之间的相互作用构成了基因调控网络，进而实现了细胞内多样的生物功能，并在人类生命过程中发挥着重要作用。近年来高通量技术的飞快发展带来了大规模的多组学数据，如何有效地从这些数据中提取有用信息，用于构建和分析基因调控网络，对于揭示相关功能机制有着重要意义。在本项目中，我们将聚焦于通过集成多组学数据来对基因调控网络进行建模。首先，基于网络分解和合成交替的策略，我们拟提出一种局部贝叶斯网络方法，对多组学数据内部的有向调控网络进行构建；其次，为了降低有标签数据的不足对有监督学习方法的影响，我们将提出基于强化学习模型的多组学数据内部无向关系网络的构建方法；最后，我们拟利用网络阻尼传播方法，构建表观遗传修饰调控网络。本项目的研究成果将有助于帮助生物学家或医学家们更直观地分析基因调控网络中隐含的复杂调控规律，为未来个性化医疗发展和新型药物设计提供技术平台支撑。

基因及其表达过程中形成的RNA或蛋白质等基因产物之间的相互作用构成了基因调控网络，进而实现了细胞内多样的生物功能，并在人类生命过程中发挥着重要作用。近年来高通量技术的飞快发展带来了大规模的多组学数据，如何有效地从这些数据中提取有用信息，用于构建和分析基因调控网络，对于揭示相关功能机制有着重要意义。在本项目中，聚焦于通过集成多组学数据来对基因调控网络进行建模。首先，基于网络分解和合成交替的策略，我们提出一种局部贝叶斯网络方法，对多组学数据内部的有向调控网络进行构建；其次，为了降低有标签数据的不足对有监督学习方法的影响，我们提出基于强化学习模型的多组学数据内部无向关系网络的构建方法；最后，我们利用网络阻尼传播方法，构建表观遗传修饰调控网络。本项目的研究成果将有助于帮助生物学家或医学家们更直观地分析基因调控网络中隐含的复杂调控规律，为未来个性化医疗发展和新型药物设计提供技术平台支撑。