化合物成药性预测与优化方法研究

Drug discovery is a hard journey, not only taking long time and high costs, but also full of high risk. There are too many factors to be considered in drug discovery, besides pharmacological activity, also including absorption, distribution, metabolism, excretion, and toxicity (ADMET), to ensure that the drug has good safety and efficacy, adequate bioavailability and half-life in the body, which is totally named as chemical druggability. The deficiency of chemical druggability is the key factor of failure in drug development, mostly related to drug ADMET properties. Therefore it is of great significance to develop methods for the prediction of chemical druggability and structural optimization to enhance its druggability. We have long been engaged in the study of chemical ADMET prediction, and have obtained a series of good results in this field. In this project, we will focus on method development for prediction and optimization of chemical druggability based on big data and artificial intelligence technology. We will first construct prediction models for drug absorption, metabolism and toxicities, next identify the key substructures to affect drug ADMET properties, and then develop substructural substitution methods to optimize chemical druggability. Finally, taking nuclear receptors (NR) as examples, we will perform experimental validation on druggability prediction and structural optimization for newly discovered NR hits. The developed methodology will be added into our web server admetSAR (http://lmmd.ecust.edu.cn/admetsar1/).

药物发现是一趟十分艰难的旅程，不但代价大、周期长，而且风险高。其中需要考虑的因素太多，除了药理活性外，更要综合考虑化合物在体内的吸收、分布、代谢、排泄及毒性（ADMET），以保证药物在体内具有良好的安全性和有效性、足够的生物利用度和半衰期，此即成药性。成药性不足已是药物开发失败的关键因素，因此开展化合物成药性预测，进而通过结构优化来改善化合物成药性具有重要的研究意义。我们实验室长期从事化合物ADMET预测研究，并已取得一系列的研究成果。本课题将在前期研究基础上，基于大数据和人工智能方法，重点针对药物吸收、代谢和毒性，构建新的预测模型，并识别影响药物ADMET性质的关键子结构；然后发展子结构替代等方法，对化合物基于成药性进行结构优化；最后以我们已针对核受体获得的一系列高活性苗头化合物为例，进行基于成药性的预测和结构优化实验验证。本研究发展的方法将提供在线服务。

化合物成药性主要涉及化合物在体内的吸收、分布、代谢、排泄过程及毒性（合称ADMET），是药物开发成败的关键因素。本项目围绕化合物成药性，开展了一系列研究。首先基于已知实验数据，采用人工智能技术，构建了40余种ADMET预测模型；其次发展了系统评价化合物成药性的打分函数ADMET-score和DBPP-Predictor；第三我们发展了基于骨架跃迁技术的成药性优化方法ADMETopt，并进一步结合匹配分子对规则，将其升级为ADMETopt2；我们还以核受体为例，对上述方法进行了实验验证，获得了具有更好成药性的活性化合物6个；最后，我们将已有的在线预测系统admetSAR升级到了2.0版，将项目实施过程中发展的各种ADMET预测模型陆续加入系统中，并新增了ADMET优化模块ADMETopt，获得广大用户的认可。.本项目共发表论文30篇，其中SCI收录论文29篇；申请并获得计算机软件著作权3项；申请中国发明专利2项，其中1项已获授权；获得高活性核受体配体21个，其中成药性优化后配体6个；在国内外学术会议做邀请报告9次；指导毕业博士4人，硕士5人。圆满完成了项目预定的各项任务指标。