基于多层次注释信号通路网络预测药物-癌症关系的新方法研究

Recently the research focus of drug-cancer prediction lies mostly in single or several abnormal genes. However, studies incorporating multi-level annotation signaling network to discover abnormal gene changes during drug-cancer prediction are scanty. In previous works, our group have constructed human phosphorylation mediated signaling network, and cancer-related protein variations and differential expression databases; have explored predictive algorithms for drug-disease relationships using the unilateral Pearson test and the modified Stouffer test with gene expression data. Based on these results, this project would systematically propose new ideas and new methods for "drug-cancer" prediction. First, abnormal genes will be obtained from drug treated gene expression data and cancer related high-throughput sequencing data. Then, a signaling pathway network annotated with protein variation, expression information mediated by abnormal genes will be constructed, and key genes and network modules will be extracted from this network. We will build a model to predict "drug-cancer" relations based on key genes and network modules. This study may find some key genes and network modules between drug and cancer, providing a theoretical basis for tumor network medicine research.

目前药物-癌症预测研究更多的关注单个或几个异常基因的变化，然而从多层次注释的信号通路网络的角度考虑异常基因改变的研究相对匮乏。本课题组在前期组学数据分析研究中，已构建了人磷酸化介导的信号传递网络数据库，以及癌症相关蛋白质差异表达和突变的注释数据库；并且使用基因表达谱数据，通过单边皮尔逊检验和修改的Stouffer检验，探索药物和疾病关系的预测算法。本项目在此基础上，拟整合药物处理后基因表达谱的数据特征，及癌症相关的高通量测序数据特征，获取异常基因；重构异常基因介导的多层次注释信号通路网络，挖掘网络关键基因和模块；通过网络药物预测模型，系统的提出药物-癌症预测的新思路和新方法，并在方法建立、评估与应用方面展开研究。该项目可能发现药物和癌症之间关键的基因和网络模块，为肿瘤网络药物研究提供理论基础。

本项目基于基因表达和药敏实验数据，对批量药物构建了可靠的敏感性预测模型，为肿瘤药物敏感性预测领域提供新的策略和方法。进一步的，通过整合免疫和甲基化两个层面的相关基因，为早期肺腺癌患者的个性化治疗提供了可靠的模型和候选生物标记，并发现TYMS在肺癌病人中的表达可以作为小分子药物培美曲塞用药的一个潜在指标。从多组学角度系统阐释乙型肝炎病毒引起的宿主基因表达变化，发现了宿主特异性靶标基因SIRT6，并验证了SIRT6特异性变构激动剂MDL800作为早期治疗HBV感染的先导化合物的潜力，为促进针对 HBV 感染的治疗开发开辟了一条新途径。总之，该项目从多个角度研究和发现疾病的潜在药物靶点。