药物靶标关联关系的预测方法研究

药物靶标关联关系研究是当前生物信息学中的重要研究内容之一。基于传统生物医学实验在药物靶标关联关系研究上的效率限制，通过计算的方法来预测药物靶标之间的关联关系，对加快药物的研发过程和降低疾病的治疗成本具有重要的意义。本项目提出了药物靶标关联关系预测方法研究这一课题。从药物靶标关联关系预测的计算问题出发，通过已有的公开数据集构建药物靶标关联关系和特征的基础数据集。基于此基础数据集，研究基于注意力机制的药物靶标关联关系预测方法。此外，miRNA作为新的靶向小分子，药物miRNA的关联关系预测也是本项目研究的重要内容，通过基于矩阵分解的模型来预测药物miRNA关联关系。药物的渗透性对药物能否靶向靶标有着关键的影响，故本项目进一步研究了面向中枢神经系统的药物对血脑屏障的渗透性。最终通过使用组件化的技术，开发一个构建和评估药物靶标关联关系预测方法的软件平台，为同行研究人员和生物医学实验提供支持。

The research of drug-target interaction is one of the important research topics in bioinformatics. Based on the efficiency limitations of traditional biomedical experiments in the study of drug-target interaction, predicting the drug-target interaction based on computational methods is of great significance to speed up the process of drug development and reduce the cost of disease treatment. This project proposes the topic of research on prediction methods of drug-target interaction. Starting from the calculation problem of drug-target interaction prediction, the benchmark datasets of drug-target interactions and features are constructed from existing public databases. Based on these basic datasets, the prediction method of drug-target interactions by attention mechanism is studied. In addition, miRNA as a new targeted small molecule, the prediction of the drug-miRNA associations is also an important part of this research. The model based on matrix decomposition is used to predict the drug-miRNA associations. The drugs permeability has a key impact on whether drugs can target its target, so for the drugs to treat the diseases of central nervous system, this project further studies their permeability of the blood-brain barrier. Finally, through the use of componentized technology, a software platform for constructing and evaluating drug-target interaction prediction methods is developed to support fellow researchers and biomedical experiments.