基于网络的药物－靶标相互作用预测的模型研究

项目名称： 基于网络的药物－靶标相互作用预测的模型研究

项目编号： No.61472205

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 计算机科学学科

项目作者： 曾坚阳

作者单位： 清华大学

项目金额： 83万元

中文摘要： 随着高通量实验数据测量技术的进步和系统生物学的发展，基于网络的药物－靶标相互作用预测方法逐渐成为药物重新定位和药物开发的一个新模式。本项目研究该新兴领域若干尚未解决的技术难题。首先，为了描述多维的药物－靶标相互作用关系，本项目提出基于局限型波兹曼模型的网络预测模型。其次，为了系统地整合不同平台的大规模数据，本项目提出采用多模式深度信念网模型来捕捉药物－靶标之间的深层次关系。此外，为了有效地将已知的分子结构信息结合到药物－靶标相互作用网络中，本项目提出改进的深度学习算法来寻找药物的新用途。最后，为了解决针对多个靶标的药物重新定位问题，本项目提出基于图分割的方法来预测未知的药物－靶标相互作用。本项目有关药物－靶标相互作用网络模型的研究能够大力促进药物重新定位和药物开发的发展。

中文关键词： 计算生物学；药物信息学；应用机器学习；蛋白质结构；生物信息

英文摘要： In silico prediction of drug-target interactions plays an important role towards identifying and developing new uses of existing or abandoned drugs. Network-based approaches have recently become a popular tool for discovering new drug-target interactions. Unfortunately, most of existing network-based approaches still suffer from several drawbacks. In this project, we aim to address several important problems in this new field. First, we propose to formulate drug-target interactions into a multidimensional network, and apply restricted Boltzmann machines（RBMs）to predict unknown types of drug-target interactions. This RBM-based approach can not only improve the accuracy of drug-target interaction prediction, but also extend our understanding of the molecular basis of drug action. Second, to systematically integrate large-scale multiple-platform data, we propose to use multimodal deep belief nets to predict unknown drug-target interactions. Third, to effectively integrate known molecular structural information into drug-target interaction networks, we propose to use modified versions of RBMs and multimodal deep belief nets to discover new drug-target interactions. Finally, we formulate the prediction problem against multiple targets into an energy minimization framework and apply a graph cut approach to identify a drug that can interact with multiple targets. The above models developed in this project will significantly advance the fields of drug repositioning and drug discovery.

英文关键词： Computational Biology;Pharmaceutical Informatics;Applied Machine Learning;Protein Structure;Bioinformatics