计算机辅助极光激酶A和极光激酶B抑制剂的研究

Aurora-A and -B kinases are important targets for anticancer drug discovery. These two homologous Aurora kinases are very similar in sequences and structures, however, they have different functionalities in organisms, and sometimes they are overexpressed in different types of cancer. In order to cure different types of cancer, it is necessary to find new highly active and selective inhibitors to Aurora-A or -B kinase; it is also important to understand the mechanism why an inhibitor can selectively interact with Aurora-A or -B kinase. In this project, firstly we will establish comprehensive databases for inhibitors of Aurora-A and -B kinases, respectively. Based on them, the classification models for distinguishing high and low active inhibitors and the quantitative models for predicting the bioactivity values of inhibitors will be built using Self-Organizing Map (SOM) and Support Vector Machine (SVM) methods, etc. With these models, the compounds virtual screening calculations will be performed (for finding compounds with high predicted bioactivity values) on the databases such as ZINC, PUBCHEM, MDDR, and Traditional Chinese Medicine Database, etc. For the compounds obtained from the above step, their ADMET properties will be computed. Next, the compounds with high predicted bioactivity values and good ADMET properties, are to be further studied individually by calculating the molecular interaction between each compound and Aurora-A (or-B) kinase using docking and Molecular Dynamics (MD) simulation. Then the novelty evaluations for the scaffolds of compounds will be performed. Finally, the compounds supposed to be potential and selective inhibitors of Aurora-A or -B kinase, will be tested by bioassay experimentally. In all, one aim of this project is to find highly active and selective inhibitors of Aurora-A or -B kinase with novel scaffolds. In addition, the mechanism for the selectivity of inhibitors will be investigated by computing their substructures, classifying them according to their scaffolds, and by computing the detail interactions between each inhibitor and Aurora-A or -B kinase, respectively. The study in this project will be helpful for anticancer drug design on the basis of Aurora-A and -B kinases.

极光激酶A和B是重要的抗癌药物设计靶标，二者之间结构相似，但在人体的功能不同、在不同癌症中的过表达程度不尽相同。为更好地治愈不同的癌症，有必要寻找新的高活性高选择性抑制剂，并研究其对极光激酶A和B的选择性作用机理。本项目拟建立全面的极光激酶A和B抑制剂数据库，采用自组织神经网络和支持向量机等算法分别建立抑制剂高活性与低活性分类模型、抑制剂活性定量预测模型。利用这些模型对ZINC、中药数据库等虚拟筛选；评价ADMET性质；进行化合物与极光激酶分子对接和动力学模拟；评价化合物骨架的新颖性；对以上得到的（高预测活性、良好ADMET性质、与极光激酶强相互作用、且有新型骨架）化合物进行活性测试。突破已有分子骨架，找到新型的高活性高选择性的极光激酶抑制剂。通过对抑制剂子结构计算、骨架结构分类、抑制剂与极光激酶相互作用研究，阐明其对极光激酶A和B的选择性作用机理，为基于极光激酶的抗癌药物设计提供新思路。

本项目建立了全面的极光激酶抑制剂的生物活性数据库。采用多种机器学习算法，系统全面地建立了一系列极光激酶抑制剂生物活性预测的模型，包括极光激酶A和B抑制剂的活性定性分类模型和定量预测模型，以及极光激酶A和B的选择性抑制分类模型；采用分子形状、静电相似性筛选与极光激酶抑制剂活性预测的机器学习模型预测相结合的基于配体的虚拟筛选，对总数超过500万个小分子的大型数据库进行逐级漏斗式虚拟筛选，得到24个极光激酶的潜在高活性的苗头化合物，经Mobility Shift Assay体外酶活实验测试，发现六个新型的极光激酶抑制剂，其中三个双效抑制剂（两个化合物的抑制活性在纳摩尔级别上），一个选择性抑制极光激酶A的化合物，以及两个选择性抑制极光激酶B的化合物。本项目的研究成果对进一步基于极光激酶的药物开发具有重要的意义。在此基础上还完成了以下方面工作：抗癌药物靶标人类表皮生长因子受体-1和受体-2、Polo样激酶1（PLK1）及其抑制剂生物活性的计算预测研究；抗丙型肝炎病毒药物靶标NS3/4A蛋白酶及其抑制剂的计算预测研究；抗炎药物靶标及其抑制剂活性的计算预测研究，包括人源5-脂氧合酶和膜结合型前列腺素E2合酶I型（mPGES-1）；抗艾滋病药物靶标HIV-1 整合酶及其抑制剂活性的计算预测研究；抗疟疾药物靶标恶性疟原虫葡萄糖-6-磷酸脱氢酶（PfG6PD）及二氢乳清酸脱氢酶（Pf DHODH）与其抑制剂活性的计算预测研究；抗病毒药物靶标H1N1抑制剂，以及抗阿兹海默症靶标Beta 分泌酶（BACE1）抑制剂的生物活性计算预测研究；中药寒热属性的分类预测研究。