药物诱导靶器官毒性预测模型的构建及毒性分子结构特征的分析研究

The chemical-induced target organ toxicities, including cardiotoxicity, hepatotoxicity, pulmonary toxicity and nephrotoxicity, are most common causes of drug R&D failure and market withdrawal. In order to reduce the risk of late drug development, early evaluation of target organ toxicities at the stages of drug design and discovery is very important. At present, the commercial toxicity prediction softwares mainly focus on these toxicities that are easy to be found in preclinical studies, but seldom predict and evaluate the target organ toxicities. Thus, in this study, the SVM and Bayesian and recursive partitioning classifier methods will be used to develop the target organ toxicities prediction models, including cardiotoxicity, hepatotoxicity, pulmonary toxicity and nephrotoxicity. The classical toxicity test data based on mouse will be used to measure the reliability of these established models. The developed prediction models for target organ toxicities will be shared on internet platform. The molecular descriptors related to these target organ toxicities and some toxic fragments will be carefully analyzed, which could provide some theoretical guidance for medicinal chemists in the design of new candidate drugs and lead optimization. In summary, the implementation of this project will make certain contributions to the early rapid screening and prediction of these target organ toxicities and optimization and modification of target organ toxic molecules.

心脏毒性、肝毒性、肺毒性及肾脏毒性等靶器官毒性是导致药物开发失败和从市场上撤回最为常见的原因。为降低药物后期研发的风险，将靶器官毒性评价提前到药物设计和发现阶段非常重要。目前商业化的毒性预测软件着重于预测临床前研究比较容易发现的毒性，而很少对药物分子的靶器官毒性进行预测评价。基于此，本课题针对心脏毒性、肝脏毒性、肺毒性和肾脏毒性等靶器官毒性将开展以下研究：采用SVM、贝叶斯及递归分类决策树方法构建靶器官毒性计算机辅助预测模型；利用经典的老鼠毒性试验数据检验模型的可靠性；将所构建的靶器官毒性计算机预测模型共享互联网平台；深入分析与心脏毒性、肝毒性、肺毒性及肾脏毒性相关的分子描述符、毒性分子片段及毒性分子结构特征，可为候选化合物分子优化提供一定的指导性意见。总之，本课题的研究将能够为心脏毒性、肝毒性、肺毒性及肾脏毒性等靶器官毒性的早期快速筛选评价及毒性分子优化修饰做出一定的贡献。

心脏毒性、肝毒性、肺毒性及肾脏毒性等靶器官毒性是导致药物开发失败和从市场上撤回 最为常见的原因。为降低药物后期研发的风险，将靶器官毒性评价提前到药物设计和发现阶段非常重要。本项目采用SVM、贝叶斯及递归分类决策树方法对心脏毒性、肝脏毒性、肺毒性和肾 脏毒性等靶器官毒性展开预测模型及毒性分子特征的研究：（1）采用SVM、贝叶斯及递归分类决策树方法成功构建心脏毒性计算机预测模型，同时分析发现与心脏毒性相关的13个分子描述符及心脏毒性分子片段。（2）成功构建了肝毒性的贝叶斯和递归决策树的计算机预测模型。发现18个分子描述符及肝毒性分子片段与肝毒性密切相关。（3）成功构建肺毒性的SVM、贝叶斯和递归决策树的计算机预测模型。分析得到了与肺毒性相关的6个分子描述符及肺毒性分子片段。（4）成功构建了药物诱导肾脏毒性的SVM、贝叶斯和递归决策树的计算机预测模型。分析发现了7个描述符及肾脏毒性分子片段对识别肾脏毒性化合物非常重要。（5）将所构建的靶器官毒性计算机预测模型已共享互联网平台（https://sky.nwnu.edu.cn/2020/0314/c1243a138692/page.htm）。总之，本项目的研究成果能够为心脏毒性、肝毒性、肺毒性及肾脏毒性等靶器官毒性的早期快速筛选评价及毒性分子优化修饰做出一定的贡献。

内容获取失败，请点击重试