靶标蛋白可药性构象的预测方法发展及其在药物设计中的应用

It is well-known that the protein structure in human body constantly changes. The protein conformations that can regulate the protein function for therapeutic purpose via binding or induced-fitting drug molecules are called druggable conformations. Different drugs may be developed based on different druggable conformations of a specific target protein. For instance, the hypertension drug Metoprololthe and the heart failure drug Prenalterol are actually the antagonist and agonist of β1-adrenergic receptor, respectively. To meet the great challenge of experimentally determining dynamic conformations of target protein, this proposal aims at developing a computational approach for predicting druggable conformation. Firstly, the complete space of low-energy conformations of a protein will be created either by theoretical modeling or/and enhanced sampling modeling methods based protein sequence and PDB structures. Secondly, classification of the created protein conformations will be performed to produce a series of conformation clusters. Lastly the druggable conformation will be identified based on the clustering by our NUMD method and the prediction model of druggable binding site. For validation and application of the prediction approach of druggable protein conformations, two target proteins, namely VEGFR and RR (Ribonucleotide Reductase), will be used as examples to testify the accuracy and reliability of the prediction result, meanwhile, to obtain novel inhibitors of VEGFR and RR as lead compounds for drugs to cure cancer, retinal degenerative diseases and hepatitis B. The prediction method developed by this project will not only improve the success rate of the structure based drug design approach, but also provide new idea and powerful theoretical tool for new conformation and mechanism based drug research.

蛋白质的结构在体内是动态变化的。我们将可与药物分子结合/诱导契合，进而调节蛋白质功能达到治疗目的的构象称为可药性构象。基于不同的可药性构象可开发出不同药物，如高血压药物美托洛尔及心衰药物普瑞特罗分别是β1肾上腺素受体的拮抗剂和激动剂。受制于可药性构象实验测定的巨大挑战，本项目将根据蛋白序列或晶体结构信息，结合理论模建和增强型取样模拟等方法得到较完备的低能构象空间。然后对构象空间进行聚类分析，得到一系列合理构象。最后通过我们开发的NUMD模拟及药物结合位点识别等方法遴选可药性构象。同时选择代表性靶标体系（VEGFR及RR）开展方法验证和应用研究，在优化可药性构象预测方法的同时，得到作用机制新颖的治疗肿瘤、视网膜退行性疾病及乙肝的VEGFR及RR抑制剂先导化合物。本项目开发的可药性构象预测方法将有利于提高药物设计的成功率，并为基于新结构、新机制的创新药物研究提供新的思路和可靠的理论工具。

生物体内的蛋白质往往同时存在多个不同的构象，不同构象上的配体结合空腔也可能显著不同，能和小分子结合并能调控蛋白质功能从而实现疾病治疗的构象称为蛋白质的可药性构象。目前已经有蛋白质可药性口袋及对应的可药性构象预测方法，但大多针对静态或小规模构象变化情况。本项目发展了一种基于大规模分子动力学模拟轨迹文件的可药性口袋预测方法D3Pockets，可通过分析口袋的稳定性、连续性及相关性来判断口袋的成药性，进而获得相应的靶标蛋白可药性构象，可为化合物发现与结构优化提供可药性的构象。利用发展的可药性位点及构象预测等多种分子模拟及药物设计方法，针对与肿瘤、血管增生等相关靶标蛋白，如VEGFR2（KDR）、TRIP13、TG2、RR等蛋白质，开展了分子动力学模拟，预测研究了可药性构象及口袋，开展了虚拟筛选及药物设计与合成研究，并进一步与药理研究工作者合作开展蛋白、细胞及小鼠体内活性评价试验，共发现了具有抗肿瘤、视网膜病变等体内活性的新化合物9个。这些结果不仅说明了我们发展的理论方法及程序具有较好的实用性，能显著提高先导化合物的发现效率，而且获得了具有较好成药性的新颖活性化合物，为肿瘤、血管增生等疾病的治疗提供新的靶标、结合位点及新化合物结构信息，值得进一步研究开发。项目实施期间，部分研究成果已发表研究论文10篇，提交专利申请3件、软件著作权1件。