癌症基因组测序分析鉴定驱动基因及其路径的方法研究

Cancers are caused by the accumulation of genomic alterations. Driver mutations are required for the cancer phenotype, whereas passenger mutations are irrelevant to tumor development and accumulate through DNA replication. A major challenge facing the field of cancer genome sequencing is to identify cancer genes with mutations that drive the cancer phenotype. An impetus for the identification of driver mutations is the potential for therapies targeted against the products of these aberrant genomic alterations. To overcome this challenge, we recently developed a novel, innovative tool to identify driver genes and pathways (DrGaP) in cancer genome sequencing studies (Am J Hum Genet. 2013; 93:439-51). This application proposes to develop and improve our DrGaP by two important extensions: 1) functional consequence of different non-silent mutations and 2) tumor mutation heterogeneity. After successful methodological derivation and simulations, we will develop an efficient program to implement the proposed statistical approaches and release a newer version of DrGaP. Finally, we will apply our newly developed tool to the analysis of exome sequencing data from The Cancer Genome Atlas (TCGA). We expect to identify a complete array of driver genes and pathways altered in various cancers, many of which will be novel mutations. Such results are highly likely to provide new biomarkers for prognosis and targets for anti-cancer therapy and will therefore have a substantial impact on public health.

所有癌症都源自于体内肿瘤细胞的DNA序列异常，但并不是癌症基因组中所有体细胞突变都参与了癌症发展，只有极少数突变是促使正常细胞向癌细胞发展所必需的，即驱动突变。这些突变有望成为药物靶点和应用于个性化的基因靶向癌症治疗。目前癌症基因组测序领域面临的一个重大挑战就是如何从癌症基因组中存在的大量突变中筛选出驱动基因。本研究是基于我们最近研制的鉴定驱动基因和代谢路径的统计算法和生物信息学工具DrGaP（Driver Genes and Pathways）的基础上，针对两个潜在问题做重要的改进：非沉默突变的不同功能性后果和癌突变的异质性。开发其相应的分析驱动突变的生物信息学软件和相应的技术分析平台，并将该平台应用于分析TCGA(The Cancer Genome Atlas)大规模外显子测序数据，挖掘各种癌症的驱动基因。本项目的完成将对癌症基因组分析与改善公共健康的目标产生重大影响。

所有癌症都源自于体内肿瘤细胞的DNA序列异常，但并不是癌症基因组中所有体细胞突变都参与了癌症发展，只有极少数突变是促使正常细胞向癌细胞发展所必需的，即驱动突变。这些突变有望成为药物靶点和应用于个性化的基因靶向癌症治疗。目前癌症基因组测序领域面临的一个重大挑战就是如何从癌症基因组中存在的大量突变中筛选出若干驱动基因。本研究基于我们前期研制的鉴定驱动基因和代谢路径的统计算法和生物信息学工具DrGaP（Driver Genes and Pathways），发展了一种基于机器学习和多种统计学原理的癌症驱动基因鉴定方法，即DriverML（https://github.com/HelloYiHan/DriverML）。该方法考虑了各个样本、基因和突变类型的背景变异率、多种突变类型对蛋白质功能的影响，采用机器学习和得分检验来鉴定驱动基因，具有高稳健性，广泛适用于各类型癌症数据。我们将DriverML应用于TCGA中10种不同肿瘤突变数据（GBM，OV，BRCA，LUSC，AML，THCA，LUAD，BLCA，HNSC，LGG）中，根据4种鉴定肿瘤驱动基因的基准方法（CGC，Mut-driver，HiConf和与其他方法的重叠），与现有鉴定肿瘤驱动基因的其他16种方法（ActiveDriver, Dendrix, MDPFinder, Simon, NetBox, OncodriverFM, MutSigCV, MEMo, CoMDP, DawnRank, DriverNet, e-Driver, iPAC, MSEA, OncodriveCLUST和DrGaP）进行比较。基于大规模和无偏的比较发现，DriverML鉴定的驱动基因的表现最好，在敏感性和特异性间达到了较好的平衡，既能检测到较多数量的驱动基因，又能保持较低的假阳性。本方法及其开发的计算机软件对于寻找癌症治疗的潜在位点和抗癌药物开发都具有重要意义。