miRNA与疾病相互作用预测及其在非小细胞肺癌分析中的应用研究

The abnormal regulation of miRNA is closely related to the disease. It is of great significance to explore the potential relationship between them for understanding the pathogenesis, diagnosis and treatment of the disease. The existing calculation methods have some problems, such as poor prediction performance of new diseases (or miRNA), single information discovered and lack of specific application. Under the background of big data, this project is based on multiple perspectives to identify disease-related miRNA, the specific types of association and Co-regulation module: (1) construct a deep learning model based on Autoencoders for unsupervised feature learning of massive RNA sequences to identify the target relationship between miRNA and mRNA; (2) using the information of miRNA target gene, sequence and function to construct the similarity network, and based on the link propagation algorithm to predict the disease-related miRNA; (3) to construct the hybrid restricted Boltzmann Machine model to further predict the specific types of miRNA- disease association; (4) considering the influence of the interaction between lncRNA,mRNA and miRNA on the disease, the co-regulation module related to the disease is mined based on the local information of nodes and the strategy of community fusion. This study is helpful to provide a new biomarker for the non-invasive diagnosis and prognosis of complex diseases such as non-small cell lung cancer and provide new clues for its treatment.

miRNA的异常调控与疾病密切相关，探索两者潜在关系对于疾病发病机制的理解和诊断治疗具有重要意义。现有计算方法存在对新的疾病（或miRNA）预测效果不佳，发现的信息比较单一及缺少具体应用等问题。本项目在生物大数据背景下，基于多视角，以识别疾病相关的miRNA、具体关联类型、共调控模块为主线展开研究：（1）构建基于自编码的深度学习模型对海量RNA序列进行无监督特征学习，识别miRNA与mRNA的靶向关系；（2）利用miRNA靶基因、序列、功能等信息构建相似性网络，基于边传播算法预测疾病相关的miRNA；（3）构建混合受限玻尔兹模型对miRNA-疾病关联的具体类型进一步预测；（4）考虑lncRNA、mRNA与miRNA相互作用对疾病的影响，基于节点局部信息和社区融合策略对疾病相关的共调控模块进行挖掘。本项目有助于为非小细胞肺癌等复杂疾病的非侵入式诊断预后提供新的标记物，为其治疗提供新线索。

人类复杂疾病的发生与microRNA（miRNA）异常表达密切相关。探索miRNA-疾病潜在关系对于疾病发病机制的理解和诊断治疗具有重要意义。本项目在生物大数据背景下，基于多视角对疾病与miRNA相互作用关系、致病模块挖掘等方面进行了深入探讨，研究内容已经按计划完成。.项目组根据研究目标，结合序列、结构、测序表达等多源数据采用深度学习理论建立miRNA及其靶标的识别模型，探索miRNA与靶标的调控机制，构建miRNA多重相似性网络，结合miRNA与疾病关联网络和疾病语义相似性网络构建全局的miRNA和疾病异构网络，建立基于复杂网络的miRNA-疾病关联预测模型，探索基因调控关系和疾病特异性共调控模块。经过研究，本课题组构建了可靠的miRNA与疾病关联权重网络，设计了精确的miRNA-疾病关联预测及功能挖掘模型，提出了miRNA-药物小分子相互作用预测方法、局部和全局拓扑关系相结合的基因调控网络推断方法、可解释序列的增强子及其强度预测方法，疾病与微生物等其他因素关系等问题的研究方法。该研究成果将有助于发现新的miRNA及其调控功能，从miRNA层面揭示疾病的发生发展机制，为相关疾病的诊断治疗、药物研发提供数据支撑和导向。