项目名称: 基于几何算法与机器学习的反向配体结合位点预测
项目编号: No.11301286
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 数理科学和化学
项目作者: 王奎
作者单位: 南开大学
项目金额: 22万元
中文摘要: 蛋白质-配体结合位点的预测,是生物信息学中的重要问题。广泛应用于蛋白质功能注释,虚拟药物筛选,合理化药物设计中。反向配体结合位点的预测不同于一般蛋白质-配体结合位点的预测。一般蛋白质-配体结合位点预测中给定蛋白质靶点,预测哪些配体与之结合。而反向配体结合位点预测,是指对特定的配体,预测与之结合的蛋白质,进而对该配体的功能与作用进行推断。反向配体结合位点预测对于理解蛋白质-配体结合机理,药物副作用的发现等方面有着重要的作用。 本项目研究的主要内容和创新点包括:提出从反向配体结合位点预测到药物副作用推测的系统研究方案;将机器学习与几何算法相结合进行反向配体结合位点的预测;技术上,将统计深度函数,gamma可达半径等几何特征引入到蛋白质-配体结构的几何计算中。这些创新将有效地提高反向配体结合位点的预测精度,可作为蛋白质-配体结合模式分析与重大疾病相关药物副作用研究的基础。
中文关键词: 反向蛋白质-配体结合位点;药物靶点;蛋白质-药物作用数据库;机器学习;几何算法
英文摘要: The protein-ligand binding site prediction plays an important role in bioinformatics. It was applied in many fields, such as protein function annotation, virtual screening of drugs, raional drug design etc. The inverse ligand binding prediction is used to find off-targets of a given ligand based on its known interactions with the therapeutic target, then deduced the side effect of ligands/drugs. It is different from finding ligands/drugs for a given protein target and is important in discovery of drug side effect. The main idea of this application is to provide a path from computational predcition to discovery of drug side-effect. We combine the machine learning model and geometric algorithm to improve the inverse ligand binding predcition. We also introduce new protein-ligand surface descriptors such as gamma radiu and statistics depth . These idea will do a great help to obtain a much better prediction. The improvement will provide a deep insight of protein-ligand binding model and the foundation of drug side effect discovery.
英文关键词: inverse protein-ligand binding site;drug-targets;protein-drug interaction database;machine learning;geometric algorithm