联合用药治疗具有提高疗效和降低毒性的作用，尤其在癌症治疗中起着重要的作用。然而，在个体药物数量和剂量组合形成的巨大搜索空间中，识别和验证的有效协同药物组合是非常有限的。在本项目中，我们将首先融合多源数据基于网络药理学预测协同药物组合，然后系统地设计基于剂量效应的协同作用量化方法。算法的设计主要按照以下四个思路展开：1）引入张量的概念，通过整合药物和细胞系的多个异质数据源，建立预测不同细胞系药物组合的框架；2）利用大规模成对基因表达和药物反应数据,采用基于半监督聚类方法来识别药物联合模块；3）考虑到利用实验获得剂量-效应关系的费用是极其昂贵的，利用矩阵分解法计算预测得到药物对的全剂量效应矩阵；4）基于剂量效应关系数据，设计计算参考模型量化协同作用并确定其协同和拮抗剂量区域。本课题的完成不仅将促进协同药物组合的发现和研究，而且将为基于联合用药的癌症治疗提供理论基础。

Combination therapy, offering increased therapeutic efficacy and toxicity reduction, plays an important role in treating cancers. Yet, among the huge searching space formed by the number of individual drugs and dose combinations, the effective synergistic drug combinations which are already identified and validated are limited. In this project, we will first fuse the multi-source data to predict synergistic drug combination based on network pharmacology. Then, methods of quantifying synergy based on drug dose responses are systematically developed. The development of the proposed methods mainly consists of the following four aspects: 1) by introducing the concept of tensor, we could establish a framework for computing drug combinations of different cell lines by integrating multiple heterogeneous data sources of drugs and cell lines; 2) we could propose a semi-supervised clustering method to identify joint modules using large-scale gene-expression and drug-response data; 3) given that the cost of dose-response relationship obtained by experiments is extremely high, we could use matrix factorization for the prediction of the full dose response matrices of drug pairs; 4) based on dose-response relationship, we could develop reference model for quantifying synergy and identifying synergistic and antagonistic dose regions. The accomplishment of this project will not only promote the discovery and research of synergistic drug combination, but also provide the theoretical basis for the cancer treatment based on combination therapy.