基于新一代肿瘤测序数据的驱动通路发现与综合分析方法研究

With the emergence and rapid advance of next generation sequencing technology,which makes it feasible to hunt for important driver genes and driver pathways related to the origin and development of maligant cells within whole genome,the research progression on pathogenic mechanism of tumor is greatly accelerated. However,the current data mining and analysis methods are greatly challenged from very large scale, extremely complex and heterogeneous whole genome sequencing data that have been accumulated increasingly.Based on the whole genome sequencing data, exon sequencing data,DNA methylation data, microRNA sequencing data and digital gene expression profiles data, etc., an advanced template-based correlation filters, random walk algorithm, novel heuristic feature prioritization,and sparse representation based meta-sample analysis methods,etc. are well designed to recoginze a few of driver mutations and genes from a great number of passenger mutations,to classify and cluster tumor subtype, and to construct gene regulatory network. The experimental results will be integrated with protein-protein interaction network to find important driver mutations, genes and pathways.Furthermore, to achieve the goal of analyzing the evolution model of tumor progression, the balance relationship between driver mutation and passenger mutation will be studied to analyze the effects of passanger mutation on tumor progression and how the combinations of driver mutations corresponds to specific tumor subtype.This research is of great benefit to the software development of clinical tumor diagnosis and prognosis,the research and development of anticancer drug as well as personalized treatment and medicine.

新一代高通量测序技术的出现极大地加速了肿瘤致病机理的研究进程，为在全基因组范围内分析导致肿瘤发生发展的驱动突变提供技术保障，然而目前的数据挖掘方法研究却受到快速积累的超大规模的复杂异构测序数据的巨大挑战。我们的研究就是以全基因组测序、外显子测序、DNA甲基化、小RNA测序以及数字化基因表达谱等数据为基础，设计基于模板的高级相关滤波器、随机行走算法、基于启发信息的基因选择以及基于稀疏表示的元样本数据分析等方法，开展驱动突变与伴随突变识别、肿瘤分类与聚类和基因调控网络构建等的综合分析方法研究，其实验结果再与蛋白质相互作用网络结构融合以期发现肿瘤相关的重要驱动突变、驱动基因和驱动通路。通过分析驱动突变与伴随突变的制衡关系，研究伴随突变对肿瘤发展进程的影响以及驱动突变组合与特定肿瘤亚型的关系，以达到分析肿瘤发展演化模型的目的，为研制肿瘤临床诊断与预后分析软件、肿瘤药物研制及其个性化医疗提供理论基础

本课题就是以全基因组测序、外显子测序、小RNA 测序以及数字化基因表达谱等数据为基础，设计一系列算法开展驱动突变与伴随突变、驱动基因与驱动通路等识别方法研究，并进一步分析驱动基因发生的时序关系，推断肿瘤演化进程，为研制肿瘤临床诊断与预后分析软件、肿瘤药物研制及其个性化医疗提供理论基础。例如，我们提出一种结合人工鱼算法和遗传算法的混合方法来识别驱动通路，该方法能够在没有使用先验知识的前提下解决最大权重子矩阵问题；提出一种有效的基于图熵的肿瘤关键基因识别算法，该算法结合基因表达数据和基因突变数据识别肿瘤相关关键基因；设计了一个从体细胞突变数据中识别突变驱动通路以及在通路水平上推断肿瘤演化进程的框架。另外，我们还提出一种结合相位相关的滤波器方法（Phase-only Correlation, POC）与改进的粗糙模糊聚类（Rough Fuzzy Cluster，RFC）相结合的调控模块识别方法，以识别miRNA调控模块，并进一步结合mRNA、microRNA、蛋白质表达数据和驱动基因信息识别癌症相关的微小RNA。项目执行期间，共发表25篇SCI与EI检索论文；申报了名为《一种基于miRNA-基因调控模块的癌症相关MicroRNA识别方法》的专利（申请号：20181615318.5）；成功申报了2018年CCF科学技术奖自然科学二等奖，其名称为《基于超算的精准医学技术研究》；在ICIC国际智能计算会议上成功组织了四届关于复杂疾病信息学的专题研讨会。总之，我们较好地完成了本课题的研究计划与目标。