基于个体通路活性的卵巢癌亚型发现和亚型特异的抗肿瘤药物识别

Ovarian carcinoma (OvCa) is the most lethal gynecologic malignancy with high incidence, and most ovarian cancer patients eventually die of disease recurrence after postoperative platinum and taxane chemotherapy. Therefore, identifying drug responsive specific subtypes is urgently needed. High-throughput genomic, epigenomic and transcriptomic analyses reveal that there are molecular subtypes in ovarian cancer, and the activated pathways and their responses to specific pathway targeted anticancer drugs in distinct subtypes are different. Based on high-throughout omics data, there are three research subjects as follows..(1).Based on the relative ordering of gene expressions in tumor samples, a set of cross-datasets, cross-platform prognostic marker is established;.(2).With multidimensional cancer genomics data, the activities of the drug targeting pathways can be inferred using the factor graph probability model, through the analysis of molecular interaction in the signal pathway; furthermore, subtypes with different pathway activities in the high-risk group could be found;.(3).Using cell line model, effective drugs for subtypes in high risk group can be found, and the drive genes of subtype can be validated for cell experiments;.For the high risk patients with poor prognosis, the research aims at identifying effective anti-cancer drugs according to their molecular subtypes, to make the most benefit for them.

卵巢癌是女性中高发且致死率最高的癌症类型，大多数患者在接受常规铂类和紫杉皖联合化疗后仍复发死亡，因此迫切需要识别其与药物敏感性相关的亚型以进行个体化治疗。高通量基因组、表观基因组及转录组分析提示了卵巢癌中存在着不同的分子亚型，它们的通路活性及对靶向不同通路的药物反应也不尽相同。利用高通量组学数据，本课题拟：（１）基于肿瘤样本内的基因表达水平相对大小秩序关系，发展一组稳健的跨数据集、跨平台的预后标志；（２）基于卵巢癌多维组学数据，利用因子图概率模型解析信号通路内的分子相互作用、推测药物靶向通路的活性，并据此发现预后高风险组内存在的药物靶向通路活性差异的亚型；（３）利用细胞系模型，为高风险组内亚型推测出可能有效的药物，并解析亚型特异的上游驱动基因以进行细胞实验验证。本项目针对预后高风险的患者的分子特征，发掘潜在的有效药物，最大可能地使此类患者从中受益。

卵巢癌是女性中高发且致死率最高的癌症类型，大多数患者在接受常规铂类和紫杉皖联合化疗后仍复发死亡，因此迫切需要稳健识别预后风险的模型其与药物敏感性相关的亚型以进行个体化治疗。基于高通量组学数据，本课题：（1）结合人工神经网络与COX回归模型，建立了能够准确预测卵巢癌生存时间的预后分子标志和预测模型，可以达到相较于传统机器学习方法略高的预测效果(五倍交叉验证的concordance index平均值为0.72)。（2）分析了预测模型隐藏层中的基因和通路生物学特征，根据相应的通路活性区分亚型，对照细胞系用药数据发现可能的药物治疗方案。（3）通过泛癌分析，揭示了转移到同一器官的生物通路特性，提示了卵巢癌的转移风险和转移特异性。