基于分子互作网络的癌症驱动突变识别方法研究

Tumorigenesis is generally caused by DNA mutation in the genome of somatic cells. However, not all mutation lead to cancer. It is very difficult and challenging to identify cancer driver mutation from large amount of somatic mutation. Identification of cancer driver mutation is not only to reveal the mechanisms of tumorigenesis, but also help for the cancer therapy. Recently, the study of identifying cancer driver mutation, with low frequency or in non-coding region, has been the hottest and most difficult subject. Our project is to propose a computational framework to identify cancer driver mutation, with high or low frequency, and in coding or non-coding region. First, a reliable human molecular interaction network is to be reconstructed with high coverage, through integrating molecular interaction data and filtering false positive and noise in high-throughput data. Then, cancer specific network patterns with biological significance are to be extracted and analyzed, by combining the proposed human molecular network with omics data. Last, an algorithm of cancer driver mutation identification is to be designed based on the proposed human molecular network. In addition to solve high-frequent and coding problems, this algorithm is able to identify cancer driver mutation with low frequency or in non-coding region, and finally discover underlying unknown cancer driver mutation.

癌症的发生普遍认为是由体细胞内基因组中的DNA突变引起的。然而并非所有的突变都会致癌。从海量的体细胞突变中识别出致癌突变，是非常困难而具有挑战的。能够准确识别出致癌突变或癌症驱动突变，不仅能够揭示癌症的发生发展机理，而且有助于癌症治疗的研究。近年来非编码区和低频率的癌症驱动突变识别研究成为热点与难点。本项目拟构建一整套计算框架，用以识别多种癌症中，编码区和非编码区、高频率和低频率的癌症驱动突变。首先，通过对分子互作数据进行整合，去除高通量数据中的假阳性和噪声，构建覆盖度高、可靠性高的人类分子互作网络。然后，将此网络与多组学数据相结合，挖掘、分析、探索出有意义的癌症特异网络模式。最后，基于此分子互作网络，设计癌症驱动突变的识别算法，以解决编码区和非编码区、高频率和低频率的癌症驱动突变识别问题，进而探索出未知的潜在癌症驱动突变。

癌症的发生普遍认为是由体细胞内基因组中的DNA突变引起的。然而并非所有的突变都会 致癌。从海量的体细胞突变中识别出致癌突变，是非常困难而具有挑战的。能够准确识别出致 癌突变或癌症驱动突变，不仅能够揭示癌症的发生发展机理，而且有助于癌症治疗的研究。本项目主要研究：（1）通过对分子互作数据进行整合，构建人类分子互作网络。（2）基于构建的网络，结合癌症多组学数据，挖掘癌症特异网络模式。（3）设计基于癌症特异分子互作网络的癌症驱动突变识别算法。重要研究结果包括：（1）通过对30个公开分子相互作用数据源进行整合、特征提取、构建机器学习分类器、计算分子相互作用置信度得分及验证，构建出人类分子互作网络。该网络包含了约97.36%的人类蛋白编码基因及约2百万个可信度高的分子相互作用关系。（2）采用图挖掘理论与技术，开发出了基于网络拓扑结构的子图挖掘算法，并结合癌症多组学数据，在人类分子互作网络上挖掘出乳腺癌、肺癌、结肠直肠癌等多个癌症的特异网络。（3）提出了一种基于布尔网络模型的癌症驱动突变识别算法。应用该算法识别出了乳腺癌、肺癌、结肠直肠癌、卵巢癌、子宫内膜癌等多个癌症的驱动基因。课题研究对于理解基因组变异与癌症发病原理具有积极的推动作用。研究成果对于揭示生物生命过程的内部结构、癌症的治疗与药物发现具有重要的意义。