计算生物学方法发展及其在分子生物学和药物研究中的应用.docx

1、计算生物学方法发展及其在分子生物学和药物研究中的应用附件2作者姓名：张健论文题目：计算生物学方法发展及其在分子生物学和药物研究中的应用作者简介：张健，男， 1978年11月出生，2002年9月师从于中国科学院上海药物研究所蒋华良研究员，于2007年7月获博士学位。中 文 摘 要重大疾病药物靶标的发现与相关药物的开发对于改善人类生命质量起着十分重要的作用。在现代生物医学和药物开发研究中，阐明和重大疾病有密切关系的蛋白质的作用机制，构建调控网络，有针对性地设计高效、高选择性的药物分子，具有重要的科学意义和应用前景。随着信息科学和计算机技术等的不断进步，计算生物学在潜在靶标蛋白作用机制和相关作用通路

2、的研究中，起到了实验无法替代的作用，并与实验形成良好互补，在生命科学研究和创新药物发现中发挥着越来越重要的作用。与此同时，药物设计已经成为药物靶标先导化合物发现的核心技术。在本论文的第一部分，我们主要发展和完善了一些计算生物学方法和药物设计方法，包括第一章的蛋白相互作用网络预测方法发展和第二章的虚拟组合集中库程序设计。在论文的第二部分，我们阐明了多个与重大疾病密切相关的潜在药物靶标的作用机制，为实验提供了有益的信息，并发现了一些高质量的全新先导化合物。这一部分包括第三章的细菌组氨酸激酶自磷酸机理的研究及抑制剂的发现，和第四章的葡萄糖激酶变构性机制研究及调控位点的发现。蛋白质-蛋白质相互作用几乎

3、决定着生物体所有的生理功能，这方面的研究具有重要的科学价值和应用前景。近年来，蛋白质-蛋白质相互作用预测方法的研究取得了很大的进展，在多个领域发挥了重要作用。然而，当前的各类预测算法在应用的准确度和适应度上还存在较大缺陷。针对以上缺点，我们利用全新的策略和优化方法，发展了基于蛋白质序列的三联子信息表征方法，并配合我们新设计的S型可交换内核函数，使用支持向量机构造普适化的快速高精度蛋白-蛋白相互作用预测模型，相互作用预测结果优于同类方法。我们将此方法拓展到不同形式的生物体内蛋白相互作用网络预测上，取得了很好的预测结果。在待预测网络中加入少量的已知信息后，我们方法的预测范围和预测精度获得进一步提高

4、，体现了其重要的实际应用前景。在此算法的基础上，我们创建了蛋白结合对象预测服务器SPPS(组合化学为发现和优化药物先导化合物提供了广阔的前景，但它在新药开发中是一种高成本、低利用率的方法，缺乏药物化学空间的导向性。目前，组合化学发展的趋势是和合理药物设计结合起来，在合成组合化学库之前，通过分子模拟和理论计算方法合理的设计虚拟化合物的集中库，增加库中化合物的潜在活性，提高库的质量。我们发展了新的虚拟组合集中库设计方法，基于靶标受体的三维结构信息，自动搜寻最优化的具有高亲和力、高多样性，良好药物动力学性质和低毒性的小分子先导化合物结构，并编制了相应的软件程序包，该方法能够有效地提高获得先导化合物的

5、成功率。为了进一步验证这个方法，我们运用虚拟组合集中库软件，根据人亲环素酶A(CypA)的晶体结构，对CypA活性抑制剂进行结构优化。实验结果显示通过该方法设计的16个化合物全部可以与CypA结合，其中最优抑制剂的活性提升了1个数量级，证明我们发展的组合集中库设计方法在新化学实体优化方面具有较高的效率，同时也为以CypA为靶标的抗HIV病毒选择性抑制剂的进一步发展提供了优质的先导化合物基础。双组分系统是细菌体内的主要信号传导系统，核心是组氨酸激酶，广泛存在于细菌等原核生物中，迄今为止未在脊椎动物(包括人类)中发现类似系统。双组分系统参与细菌的各项生命活动，尤其是介导了细菌生长、毒力和趋化性等重

6、要行为。因此，组氨酸激酶成为当前新的潜在抗生素靶标，用于发现新一代抗耐药菌抗生素。然而，组氨酸激酶自身磷酸化的分子作用机制还不清楚，极大的延缓了其成为药物靶标的可能和针对组氨酸激酶的合理药物设计。为了阐明这个机制，我们综合应用同源蛋白质模建、分子对接、分子动力学模拟和量子化学计算等方法研究了组氨酸激酶CheA的自身磷酸化机制。根据模拟结果，我们首次提出了组氨酸激酶自磷酸化的作用模式: 接受到外界传感信号的触发分子ATP，在组氨酸激酶结合位点处进行的周期性构象变化是负责磷酸化结构域P4打开ATP盖子的原因，从而导致了接受磷酸化的结构域P1与P4结构结合，发生从ATP到P1的自磷酸化反应，启动细菌

7、向目标源的趋化性运动行为。即ATP不仅是磷酸基的供给基团，也是组氨酸激酶自磷酸化作用的激活剂。同时我们也很好阐释了当前组氨酸激酶高效抑制剂TNPATP的作用原理以及与ATP的差别。在CheA自磷酸化机制的基础上，根据组氨酸激酶的关键构象和配体-受体相互作用的药效团模型，我们应用虚拟筛选方法对超过190000的分子数据库进行搜寻，结合生物测试获得IC50在微摩尔级以下的组氨酸激酶抑制剂6个。这些抑制剂抑菌效果明显，可以破除耐药细菌的生物膜保护，并且没有类似抗菌化合物所具有的凝血毒性，具有较好的开发前景，为发现新一代抗菌素先导化合物奠定了基础。葡萄糖激酶是调节血液中葡萄糖水平的重要酶，主要分布于肝

8、细胞和胰岛细胞中。在正常生理条件下，葡萄糖激酶可以同时通过促进肝细胞合成糖原和促使胰岛细胞囊泡将存储的胰岛素释放到细胞外血液中两条途径降低过高的血糖水平。因此，葡萄糖激酶被认为是目前最有前途的发现抗糖尿病药物的潜在靶标。葡萄糖激酶具有单体自调节全局构象变化的特征，但是其变构机制及中间过程未知，无法进行合理的药物设计。这种中间过渡态在目前的实验水平上极难捕捉，而且通过实验从原子水平来获得上述机制的详细答案也是难以操作的。因此，我们利用理论预测结合实验验证的方法，来研究葡萄糖激酶全局变构机制及其激动剂的作用原理等问题。通过对葡萄糖激酶进行大规模的分子动力学模拟研究，我们对葡萄糖激酶自调节变构酶的机

9、制进行了深入系统的阐述，发现了三个葡萄糖激酶自调节变构的关键中间过渡态。根据这些过渡态，我们提出: 葡萄糖激酶从闭合的激活状态，经过三个关键过渡态，最终到达超打开的非活性状态，是其作为自身调节性变构酶的本质原因；而葡萄糖激酶的激动剂正是作用于第一个过渡态，持久激活葡萄糖激酶，从而发挥降低血糖的作用。根据这个机制，我们对葡萄糖激酶过渡态上的五个关键调节残基进行了合理的预测，这些预测被进一步的突变实验和酶动力学实验所证实。这一研究结果为阐述葡萄糖激酶在体内通过自身构象变化调节血糖水平的作用机理，以及葡萄糖激酶激动剂的合理药物设计奠定了基础。与此同时，我们还对葡萄糖激酶的催化机制进行了初步探讨，建立

10、了完整的葡萄糖激酶催化模型，预测了催化所必需的关键残基，并被突变实验所证实。根据上述结果，我们提出了葡萄糖激酶磷酸化葡萄糖的机理，为进一步研究葡萄糖激酶在体内作用机制提供了方向。关键字: 计算生物学，药物设计，蛋白质-蛋白质相互作用，虚拟组合集中库，双组分系统，葡萄糖激酶，先导化合物发现The Development of Computational Biology Methods and The Application in the Molecular Biology and Drug DiscoveryZhang JianABSTRACTDrug development and targe

11、t discovery are of great importance to improve the quality of human life. In the modern biological medicine and pharmaceutical study, investigating the mechanism of action by the critical proteins closely related to a serious disease, constructing its regulative network, and designing new drugs with

12、 high selectivity and activity are fundamental research areas with highly potential applications. With the development of computational technology and information science, it has been recognized that computational biology as a complementary tool to experiment can assist in investigating the mechanis

13、ms of protein function and the associated pathway of a target protein. Both information remains intractable through direct experimental methods. Thus, computational biology has become an essential technique in the research field of life science and drug discovery. Meanwhile, drug design has already

14、become a core technique of identifying leads for drug targets in drug development. The first part of dissertation will mainly focus on the development of a series of computational algorithms and drug design methods, including “The Prediction of Protein-Protein Interaction Network” in the chapter 1 a

15、nd “Virtual Target-Specific Compound Library Design” in chapter 2. In the second part, we will investigate the mechanism of action on several important potential target proteins, leading to the discovery of some novel lead compounds based on the discovered mechanism that will be useful for further e

16、xperimental interrogation. This part consists of “Mechanism study of Histidine Kinase Autophosphorylation in Bacteria” in chapter 3 and “Allosteric Mechanism study of Glucokinase” in chapter 4.Protein-protein interaction (PPI) is a key to understanding most of the biological processes. The research

17、field has been unprecedentedly appreciated due to its important scientific value and prospective application. Recently, prediction of PPI in computational biology has been greatly improved and plays a significant role on several fields. However, these methods still have large limitation in accuracy

18、and the scope of PPI prediction. According to such shortcoming, we proposed a novel method for PPI prediction with high accuracy and generalization. This algorithm was developed based on a new machine learning algorithmsupport vector machine (SVM) combined with a newly designed S-type exchangeable k

19、ernel function and a conjoint triad feature for describing amino acids of protein sequence. The prediction ability of our approach is better than that of other sequence-based PPI prediction methods. Remarkably, our approach has been extended to predict various PPI networks in organism, showing good

20、consistent results compared with experiments. Furthermore, a few clues from request network can effectively enhance the prediction ability of our method in accuracy and scope, which shows its important role in practical cases. Based on the algorithm, I constructed the web serverSequence based Protei

21、n Partner Search ( for public access, providing probability prediction service on searching potentially direct or indirect binding partners and the associated network of the request protein in different species, such as homo sapiens. This server has been visited more than 2500 times so far, providin

22、g guidance for experiments and being helpful to enhance the efficiency for experiments. The predicted results from SPPS server on several biological systems have already been proved by independent study laboratories. Besides, the server may help design novel therapeutics that can interrupt disease-r

23、elated protein networks rather than simply inhibiting individual target proteins.Combinatorial chemistry was previously thought to be a new efficient approach for discovering and optimizing lead compounds. But the applications of this approach in the past decade testified it to a high-cost strategy

24、with low profitability in drug discovery due to lack of direction within medicinal chemistry space. Recently, there is a new attempt that combines the combinatorial chemistry with rational drug design, which may lead to a virtual combinatorial focus library with high potential bioactivity and favora

25、ble library quality in silico before synthesis. We proposed a new method for designing novel target-focused library, which automatically optimizes small organic compounds with high binding affinity, preferable diversity, acceptable pharmacokinetic properties and low toxicity based on target 3D cryst

26、al structures. Meanwhile, a software package was developed on the basis of the new method and makes effect on increasing successful ratio of searching lead compounds by several cases. To verify the method, we used the software to improve activities of the inhibitors against cyclophilin A(CypA) based

27、 on its X-ray crystal structure, which performs an essential function in HIV-1 replication by binding specifically with the capsid domain. All these sixteen molecules are CypA binders with binding affinities ranging from 0.076 to 41.0 M, and five of them are potent CypA inhibitors with PPIase inhibi

28、tory activities of 0.25-6.43 M. Remarkably, both the binding affinity and inhibitory activity of the most potent compound increase 10 times than that of the most active compound discovered previously. The results demonstrated that our method has great efficiency in optimizing bioactivity of a new ch

29、emical entity and provides excellent selectivity inhibitor leads of CypA against HIV virus.The two-component system (TCS) signal transduction is a predominant signaling system mediated by histidine kinase in bacteria. TCSs are crucial to bacteria, for they modulate a large variety of bacterial chara

30、cteristics, such as growth regulation, virulence, and chemotaxis. In addition, TCSs are rarely found in mammalians. Therefore, histidine kinase involved in TCSs may be an attractive target for discovering novel antibiotics that may treat multiresistant bacteria. However, the mechanism of autophospho

31、rylation in histidine kinase is still unclear, which largely encumber possibility of developing histidine kinase into drug target and rationally designing new compound against histidine kinase. Taking chemotaxis protein CheA as a model of TCS, the autophosphorylation mechanism of the TCS histidine k

32、inases has been investigated by using a computational approach integrated with homology modeling, ligand-protein docking, protein-protein docking, molecular dynamics simulations and quantum chemical calculation. A mechanism of the autophosphorylation of CheA is first proposed as that ATP, which receives signal from external environment, makes conformational switch to trigger the opening of the ATP lid of P4 domain, leading