乳腺癌亚型识别的ncRNA关联信息融合方法研究

Identifying molecular subtypes of breast cancer is one of the important research directions in cancer bioinformatics. However, the pure computational models cannot be competent to achieve the accurate identification of breast cancer subtypes. This project proposes a new method for breast cancer subtype identification by utilizing the prior knowledge of ncRNA-associated gene regulation information. A complex gene regulatory network (miRNA-lncRNA-TF-mRNA) is constructed based on the integration analysis of various interatomic databases and the association analysis of breast cancer multi-omics data. A feature weighted model is designed to assign the ranking for each biological feature based on the miRNA-lncRNA-TF-mRNA regulatory network. The rankings which represent the priori knowledge are the regulatory importance of the corresponding features. A novel computational biology model is presented for clustering analysis of breast cancer multi-omics data by integrating the ncRNA-associated information. The model is an integrated analytical framework for feature extraction and multi-omics data clustering and helps to accurately identify breast cancer subtypes which meet biological significance and interpretation. Characteristic network modules and their biological functions are studied to reveal the ncRNAs’ regulatory mechanisms in each identified breast cancer subtypes. Combining with the pathological indicators and clinical data of breast cancer samples, a group of molecular biomarkers and candidate therapeutic targets are defined for each identified breast cancer subtype. The expected outcomes of this project can provide theoretical direction and important reference for assisting the design of personalized treatment therapies and targeted drug development for breast cancer patients.

乳腺癌分子亚型的识别是癌症生物信息学重要研究方向之一，然而目前基于基因组学数据的纯计算学模型难以实现乳腺癌亚型的精确识别。本项目基于ncRNA对基因调控的先验知识提出乳腺癌亚型识别的新方法。通过对大量基因调控靶标数据库的集成研究和乳腺癌多组学数据的关联分析，建立与乳腺癌密切相关的复杂基因调控(miRNA-lncRNA-TF-mRNA)网络，构造网络特征权重模型，建立特征抽取和乳腺癌多组学数据聚类的一体化分析框架，从而提出融合ncRNA调控机制信息的乳腺癌多组学数据集成聚类的新型计算生物学模型，精确识别出具有生物学意义和解释的乳腺癌亚型。研究各乳腺癌亚型内ncRNA关联的特征网络及其调控的生物学功能，结合乳腺癌样本病理学指标和临床数据进行分析验证，提出表征各乳腺癌亚型的特征分子标志物和候选治疗靶点。预期研究结果将为乳腺癌个性化医疗和靶向药物研发提供理论指导和重要参考。

乳腺癌是在分子特征上具有显著差异的异质性疾病，可以区分为不同的乳腺癌亚型，乳腺癌分子亚型的识别是癌症生物信息学重要研究方向之一。癌症基因组学大量研究证明，包括乳腺癌在内的多种癌症亚型的正确识别能够促使癌症治疗方式的革新。癌症基因组学的发展为开展相关研究提供了海量的数据支撑，提出高效的数据融合方法，从计算学角度识别具有生物学意义的乳腺癌亚型是本项目研究的核心。本项目克服前期大量的纯计算学分析模型进行癌症亚型识别的缺陷，通过融合基因调控机制的先验知识，提出乳腺癌多组学数据集成聚类的新型计算生物学模型，识别乳腺癌发生发展相关的ncRNAs与基因表达数据特有和共有功能模块：（1）通过整合乳腺癌多组学表达谱和miRNA-lncRNA-TF-mRNA的基因调控先验知识，开发了多个乳腺癌亚型识别的计算学模型，并针对识别的不同的乳腺癌亚型，进行癌症发生发展相关功能模块挖掘，提出多个表征乳腺癌亚型的特征标记分子和功能模块。（2）揭示miRNAs/lncRNAs在乳腺癌亚型识别和预后分析中的作用。研究证实集成miRNA-mRNA表达谱可以提高乳腺癌亚型识别的性能，相关乳腺癌亚型显示出差异性的生存模式特征。同时lncRNA表达谱能够促进乳腺癌亚型识别，相关lncRNA分子对乳腺癌病例预后有重要影响。（3）提出一种结构稀疏的癌症基因组数据矩阵分解框架（SSNMF），实现 L2,0范数限制性约束函数在正交非负矩阵分解（NMF）模型上的创新应用。该模型在癌症基因组数据应用上，能够有效证明鉴定出的相关基因模块的生物学功能。总之，本项目以数据融合的方法和观点挖掘乳腺癌标记基因和基因功能模块，提出表征各乳腺癌亚型的特征分子标志物和候选治疗靶点。相关研究结果将为乳腺癌个性化医疗和靶向药物研发提供理论指导和参考。本项目按研究计划顺利进行，并取得了相关研究成果，其中SCI论文7篇 ，中文期刊论文1篇，培养研究生2名，完成了预期研究的目标。